yoy.be "Why-o-Why"

Ah, zijn de liberalen daar mee bezig? Ik heb een ander idee. Wat als middelgrote bedrijven en grote bedrijven zouden worden verplicht een 'liquidatiereserve' aan te leggen? Een bedrag dat pas vrij komt als een bedrijf er mee wil stoppen of een fractie van zijn personeel aan de deur wenst te zetten.

Ik loop blijkbaar al een tijdje met het idee (2011!) maar het speelt toch maar in je hoofd. Als je weet wat bedrijven kunnen/moeten doen met sommen geld die om verschillende redenen in het bedrijf aanwezig zijn, kan je misschien zelfs verplichten om zo'n fonds lokaal te investeren. Hawel kijk, dat zou zowel de linker- als de rechter-flank dienen. En de lokale economie. (En de schatkist.) En uiteindelijk ook de mensen die jaren de onderneming dienen maar dan moeten afvloeien...

I just had a fragment of an idea. I want to write it down, just to let it go for now as I've got other things to do, and to be sure I can pick it up later exactly where I left off.

First situating what it's about: I have been reading up on WebAssembly, and to my surprise the intermediate representation is stack based (just like Java's JVM and .Net's CIL). I'm not sure why because it feels to me this makes registry assigning when constructing the effective platform-dependent instructions harder, but I may be wrong. Finding out objectively is a project on it's own, but sits on the pile 'lots of work, little gain'.

I also went through the great set of MIT 6.004 lectures by Chris Terman which really gives you a good view of 'the other side' of real assembly since it's actually born out of designing these processing units built out of silicon circuits. It prompted me to make this play thing, but again pushing that through with a real binary encoding of the instructions made it a 'lots of work, little gain' project, and I really don't have access to any kind of community that routinely handles circuit design, so it stalled there.

Before that, I read something about hyper-threading, and what it's really about. It turns out modern CPU cores actually handle two streams of incoming instructions, have a set of instruction decoding logic for each stream (and perhaps branch prediction), but share a lot of the other stuff, like the L1 cache, and especially a set of virtual registers that the logical registers the instruction stream thinks it's using is mapped on to. Mapping used registers freely over physical slots makes sense when you're making two (or more?) streams of instructions work, but it's important to know when the value in the register is no longer needed. Also if the register is only needed for just a few instructions, pipelining comes in to play and could speed up processing a great deal. But for now the CPU has to guess about all this.

When playing around with a virtual machine of my own, I instinctively made the stack grow up, since you request just another block of memory, plenty of those, and start filling it from index 0. It shows I haven't really done much effective assembler myself, as most systems have the stack grow down. What's everybody seems to have forgotten is that this is an ugly trick from old days, where you would have (very!) limited memory and use (end of) the same block for the stack, and with more work going on the stack could potentially grow into your data, or even worse your code, producing garbled output or even crashing the system. (Pac-man kill screen comes to mind, although that's technically a range overflow.) Modern systems still have stack growing down, but virtually allocate a bit of the address-space at the start of that stack-data-block to invalid memory, so stack-overflows cause a hardware exception and have the system intervene. It's a great trick for operating system (and compilers alike) to have checks and balances happen at zero cost to performance.

The consensus nowadays is that nobody writes assembler any more. It's important to know about it, it's important to have access to it, but there is so much of it, it's best left to compilers to write it for you. In the best case it may find optimizations for you you didn't even think about yourself. But this works both ways. Someone writes the compiler(s), and need to teach it about all the possible optimizations. I can imagine the CPU's instruction set manual comes in handy, but that's written by someone also, right? I hope these people talk to eachother. Somewhere. Someday. But I guess they do as with x86-64 they've kind of agreed on a single ABI... and they've also added some registers. Knowing about the virtual registers allocation going on behind the scenes, it could be that that was just raising an arbitrarily imposed limit.

So this is where I noticed a gap. When performing all kinds of optimizations and static analysis on the code when compiling, and especially with register allocation, it's already known when a register's value is no longer relevant to future instructions. What if the compiler could encode this into the instruction bits? If I were ever to pick up where I left, and have a try at a binary encoding for a hypothetical processing core, the instruction set would have bits flagging when the data in registers becomes obsolete. Since this would be a new instruction set, and I guess it's more common to need the value in a register only once, I might make it the default that a value in a register becomes obsolete by default, and you'd use a suffix in assembler to denote you want to use the value for something extra later as well.

Why, thank you. Eeuh, is it at all optimized? I took some decisions that may perform a little better than the reference implementation, but I haven't taken any time to compare to see if it actually performs any better or worse...

So MurMurHash3... Let's have a look, it's on wikipedia, so it's a thing. And the reference C implementation is in the public domain, great! Looks straight-forward enough, could boil down to a translation-job...

Rougly two hours later, got triple zero (0 errors, 0 warnings, 0 hints). Now for checking if I got it all right. Hmm, not much there, but those x32_64 match, so barring any typo's, assuming this pretty straight translate job will result in the expected behaviour for the other two functions, this should be it:

If I read about a newly discovered vulnerability related to PHP, for example this one here, I try to find out if it would apply to xxm as well.

In this case I guess there's nothing more than sending out the message, again and again, to sanitize your inputs, and poperly encode your output. Strings are never just strings. They are always an internal representation of a bit of textual data. So always think about that taking string values in, and preparing strings for output. A few weeks back I had to speak up to someone who wrote OutJSON:='{"field1":"'+value1+'"}';. Little Bobby Tables comes to mind, though I'm not sure 'JSON injection' could be so devastating as SQL injection. (And OutJSON:=JSON(['field1',value1]); is shorter!)

The other time I found out it's a really good idea to strip nastiness like EOL's (CRLF) from headers added to a response, just in case a malicious script is up to no good. Come to think of it, that's also just another case of properly sanitizing your inputs...

Is there a language that has a single word for 'the feeling of being on a shrinking island'? Anyway, this is somewhat sad to see, especially that the few most recent Delphi versions gave the impression there was a new uptake with more people getting persuaded, but it doesn't show in the curve.

I'm sorry but I feel I must react. In general I keep silent, in the hope people by themselves will know better, but as I'm getting no input what-so-ever that that is the case, I feel tempted to write something about this.

and it looks kind-of OK. To the untrained eye it looks good and may even look tempting to write more in this syntax. This is a straight-forward example of a template that works with a templating engine that no doubt has many more capabilities and features. And then I thought, learned from practice, what I typically would get asked is to not show " ()" when the PhoneExt field is empty. I would not know how to make that happen in that template syntax. That's mainly because I know nothing about the template syntax. If I look into the documentation, I might find an @if predicate to make it happen, but let's move on:

Looks roughly simlar. A little more like Delphi syntax. And in fact it is. If you know [[, ]], << and >> get translated into Context.SendHTML() and Context.Send() calls behind the scenes (full details are here), you know this code will result in the same output. Without templating engine! Streamed to the user's client! Perhaps even while the data is streaming in from the database server, in case it's a longer list, and in case there is a database server, Marco uses a memory-table for his example.

What I find important is that there's less going on between the native compiled logic and getting the data to the user launching a request. Not only a templating engine looks superfluous, this entire ORM thing is something I don't get. If it's a gigantic database model with so much tables that you clearly benefit from code-completion, then I agree, but I haven't come across something remotely close to that in web projects.

Also the HTTP-server itself is something I think that values extra attention. I've seen platforms and frameworks that offer you a wealth of capabilities and features, but hastily slapped on something that listens on TCP port for basic HTTP requests, in some cases on port 80, but more often 8080 or something else in the thousands. In real web environments, the server(s) has/have a lot more going on: load-balancing, reverse proxies, firewalling, authentication. Since we're in a Post-Snowden-era nowadays, we're all responsible to think about protecting privacy and get that HTTPS in order with the proper certification and encryption... Not to mention HTTP version 2 that's heading full steam towards being generally accepted/expected.

I can image the web-admin responsible for all that, isn't happy with your request to add this newfangled separate thing that's doing its own handling of HTTP requests. ISAPI DLL's or Apache modules play much nicer with existing IIS or Apache installations. (FastCGI is on the table, but for now xxm has SCGI available for other servers.) Even if your 'Delphi HTTP framework' of choice is specifically designed to tap your ORM of choice and offer a REST-API for your data-layer needs, it will still be one more stop along the way between the user's browser, and the delivery-setup, and the front-end, and the page-template, and the data-layer, and the database, and what the user actually needs or wants. I think of this in the postal office when there's twelve people in the queue in front of me.

I don't expect to convince much people of this way of working, but it works great for me. I remember the days with early PHP and ASP and how simple and straight-forward everything was. Knowing these work on scripting engines, I kept worrying about lost performance. This was the core reason to start xxm: employ the speed and power of the Delphi compiler to have a native library serve my websites. And it turns out that Delphi code looks quite nice between HTML to handle server-side logic, if I may say so. It took me a few years to make this happen, but I couldn't do without it any more. And people kind-of appreciate that for using this new application, all they need is their trusted browser and a URL.

If only I had known sooner! I forgot where I picked this up, but apparently if you put <label> around an <input>, typically of type checkbox or radio, browsers automatically know the label is for that control. Before, I would write my <input id="x"> first, then a <label for="x"> after. To keep code neat, I would put it on a separate line, but the EOL inbetween would not be clickable to actuate the control. This is a really minor issue, but still. Now that I know you can just write this:

Because, there are two kinds of people: those that click the box to switch a checkbox, and those that click the text right of the checbox. You might not even know that you do, but you do don't you. If you're of the latter type, it's just one of those minor frustrations, that a click on the text-label sometimes doesn't do what you expect, and you have to:

But there you have it. Heaven has great UX. Here we need to make do with what we get. (And need to make sure it's the way we like it for those bits that we have control over.)

This took more effort than I anticipated, but I'm glad I saw it through. A tiny bit of history: when I started creating a Delphi connector to access a MongoDB instance, I had to create the tools to manipulate BSON, and since I really (really!) hate long lists of overloaded methods, I created something based on Variant values. It works great for ADO, so why not. Some time later I had to manipulate JSON, unrelated to MongoDB, so I took what I got, stripped BSON-specifics, into jsonDoc. There it evolved on its own and gained some useful features.

So it was time to re-work TMongoWire, making it use jsonDoc just like any other project. There's still a bit of BSON-specific code, but it deserves a unit of its own. The downside is there's an extra unit to include in the project, but there were multiple already (jsonDoc, bsonTools, mongoWire and if needed mongoID, mongoAuth3 and mongoStream); the upside is the improved performance and features of jsonDoc, both of the current version and those the future could bring. Enough change to bump that version number:

Please let me know if the migration effort this generates is surmountable, and if I can do anything to help. In theory any references to bsonDoc, the BSON function and the IBSONDocument interface should keep working once replaced with references to respectively jsonDoc, JSON and IJSONDocument. The document object no longer implements IPersistStream, but this may have been stretching abstraction just a bit too far. We're on Delphi so plain old TStream should do, and apparently IPersistStream's Seek signature changed a tiny bit over Delphi versions, making it harder to make the code cross-version-ready.

Drats. I thought I'll try something like Clover does, start with SetParent on any top-level window you could find with GetAncestor(GetForegroundWindow,GA_ROOT) and see what it gives. There's this old default MDI child mechanic that starts working, so I thought this might just work. But it looks like over at Microsoft the Windows 10 team decided to cut some corners. The theming doesn't do it's modern stuff and reverts to Vista-style borders, and the newfangled calulator, for example, won't even take the SetParent nicely any more. Oh well. For the time I'm still on Windows 7 at work, I guess it'll come in handy there for the time being. Probably in about a week I'll know if this could work where other desktop-management tools failed or were too cumbersome or too invasive...

How does it work? Boxer starts with an empty window, and will show an icon over the top-right corner of other windows. Click that to 'box' the window. The top bar shows a tab for each boxed window. Right-click to open a menu with options to unbox or close, and drag to reorder. Multiple instances play nice with each-other and show 'boxing handles' next to eachother, the second instance labeled "A", the third "B", etc. To disable the boxing handles, right-click on the top bar to the right of all tabs.

Anyway, if you feel like having a try if you can do better, have a look at the code here. Download a binary here: Boxer.zip (209KB)

Let me tell you a story of two task-keeping-applications. Once there was a team that was struggling to keep track of the work it was doing, had done, and still had to do. It was frustrating to work on the team, hard to schedule things, almost impossible to estimate when things where going to get done. And since clients are what they are, changing requirements and additional requests would cause much more disruption than expected of a team that calmly and firmly is determined to deliver the best of market solution.

An internal brainstorming-session about what to do about it, ended in two opposing visions about what decent issue-tracking should be about, and how a system that (co-!)operates with the team members should be structured internally, should behave, and which duties it should perform either by itself, or by effect of using its rules and restrictions correctly as designed.

In an attempt to get to the best solution, two teams were formed to develop each a project based on one of the two views. An evaluation would follow determining a winner, or if would it be possible to synergize the best parts of both into a single system.

A side note: from a managerial point of view this is a very tough decision. Allocating a lot of resources to work on internal structure, takes resources away of the work that brings in revenue, and such undertakings typically risk getting frivolous or spinning out of control. But I guess it's normal that all stop rowing to help keep the boat from sinking. So, apart from being intrinsically aware of the exceptionality of this opportunity, coordinators were given instructions to keep to a strict schedule in these projects and a determined focus on delivering a workable proof-of-concept quick. Sounds like a good work-ethic to apply generally, if you ask me.

A few weeks later progress was made, and the prototypes were already being used to keep track of the issues of these new task-keeping-applications and other projects. The main difference between design visions became apparent soon enough.

One application centered around the list of work items. Care was taken the entry form was extensive enough to have fields for all of the details about a work item, its categorization, its relation to the project, an outline of the projected outcome. An overview would show the current list, possibly filtered for those items assigned to you.

The other application was centered around getting information out of the system, especially structure and relation between items. Users would enter small specific reports, and add them to the best suitable node in a tree-structure, optionally marking relations with other items over branches. The overview showed an expandable structure starting at the root items. It also could have a filter applied, but would potentially show you an entirely different structure of the same data, when using a different relation type.

Having two working prototypes, attention gradually reverted back to the serious work and some of the frustrations of before were abated. After a few months an evaluation was undertaken.

One application had rendered itself useless. The first weeks of usage, a lot of entry happened, but without consensus about categorization, the list was enormous without a clear way of grouping relevant items. Duplicate entries and ambiguous task-descriptions were unresolved, causing confusion.

The other application was doing better. It needed work, but offered a good view of what had to get done.

Enough about the story. Reality, of course, is much more bleak. As no sane manager would make such a shift in resources of a troubled organization, the 'two teams' actually stand for the existing team using an off-the-shelf something badly administered on one side, and on the other side, well, just me, toiling on something potentially better, in my off-time.

I started — like many — with a plain text-file, then a spreadsheet. Then I took the step to design a database for it, but wanted something to do entry and retrieval with roughly the same ease-of-use that a spreadsheet would offer. Working with tree-structures for some other projects, I wanted a single structure to serve as the basis to store data in. Specifically with projects and tasks — as projects tend to have sub-projects, and tasks split into sub-tasks — using branches of a tree would enable to keep this distinction conveniently vague. If you add representation for entities like users and clients into this tree-structure, and relation between nodes over branches, you've got a richness to model much more of the world, and keep track of its changes, past and future.

From there it grew slowly, over years, into what it is now: tx. First versions suffered a notoriously bad interface, but I hope that has improved. For long I was its only user, but some cooperation features have been added since. What's left for me is to keep improving tx where possible, and demonstrating it to people what it's about.

Though entry and structure is important, where a task-keeping-system can really shine is helping to keep an overview. A cleverly designed filter can limit your view to exactly these items that are relevant to you, but in tx you get the additional option of having these items display with exactly those sections of the tree-structure they have in common. Also, I personally feel the most important items on any list, as long as it may be, should be on top. So when tx displays a list of items, they get ordered using their weight, determined by a combination of factors such as task-type and current status. As a task progresses from an active state to a final state, it may move down the list or even fade from view into an archival state.

And there's much, much more I could talk about, but it's all created out of necessity and designed adhering closely to a central vision of what a task-keeping-system should be and what it should be doing for you in order to be able to depend on it.

In conclusion, and for those people that — like me — skip long stories to the last paragraph, it's so very important that information systems succeed at keeping an up-to-date model of reality, and offer you the freedom and easy to update it's view of the world. Especially so with task-keeping software you depend on to keep track of the progress you make with projects, and to help stay on top of what's important. All of that served as the basis for developing tx.

O my god I can't believe this isn't somewhere on the net already! I come across this question and ofcourse there's the default discussion about Indy and OpenSSL/LibreSSL, and Arnaud is very (very!) right about proposing to use WinHTTP. I personally would see if I could do it with MSXML2's XMLHTTP object. Contrary to what it looks like, it doesn't really have to do much about XML, and you can do really almost all of what you need out of a HTTP request with open, setRequestHeader, send and responseText or responseStream, with a much more straight-forward syntax. (If you can live with COM.)

Plus there's another pretty sound reason: back in the days, when using WinHTTP and/or VBScript (for the RegExp component...) you would make it very clear it would only work if you got Internet Explorer version 5.5 or newer installed. "How much back was this?" I hear you ask? It depends, but it's so far back that pretty much all of XML things came after, and XML being to closely related to data-processing, Microsoft needed to put MSXML*.dll deeper into the system than its web-browser. If it's tied to the MDAC or the actual core system I'm, not sure, but years of use and the success of AJAX caused it to get very many security updates. I suspect a far bit more than the DLL that handles your WinHTTP calls...

So, without further ado, do as I do and import the "Microsoft XML" type library in the most recent version you can find on your system:

On old Delphi versions (e.g. the venerable version 7) open 'Projects' on the main menu, then 'Import Type Library...', then select "Microsoft XML, v?.0 (Version ?.0)" from the list. Your system may have multiple versions installed, pick the highest version number (in my case 6.0). Uncheck "Generate Component Wrapper" as usually you don't need this, and actually have the added benefit of automatic reference counting when you use interface references to keep track of COM object instances.

On newer Delphi versions, I've noticed the import functionality has centralised, and is now under the 'Components' menu, the 'Import Component...' option and you need to select "Import a Type Library" from the first list, and then search for "Microsoft XML".

This will result in a automatically generated MSXML2_TLB.pas that you otherwise don't need to worry about. Except perhaps include it in all of your projects you'd launch HTTP requests from.
Use the CoXMLHTTP.Create class function to get a new XMLHTTP instance, and it's open and send methods to make magic happen.

There's more ofcourse. This is just the basics. Depending on what you need to do there's HTTP POST and ways to correctly encode values into the request body you pass to send, using setRequestHeader. And if you need to handle larger chuncks of data, you should check out responseBody and perhaps TOleStream declared in AxCtrls.pas.

While working on TRethinkDB, I noticed something was wrong with document re-use. My original idea was to re-use a single IJSONDocument instance to process a list of similar documents, but keep the set of keys allocated, only overwriting the values from the new document. (If you're interested about the code, search jsonDoc.pas for FLoadIndex.) Nothing an extra internal-use interface can't fix. Then I noticed that IJSONDocument.ToString could do with a revision because it wasn't using the new IJSONEnumerator, though it should.

Then I noticed the feature I once added to function JSON, where you could declare embedded documents by listing a value of '[', and closing the document with a key of ']'. The problem is that '[' could be a perfectly valid value! If it were to come via a variable, that would change the behaviour of the JSON call, and that's a big no-no. Also square brackets usually mean arrays in JSON-land, so it's perhaps confusing that I chose square brackets because you already need them in the JSON([...]) call. So I changed it into a key suffix of '{', and a closing key of '}', counting on braces being really abnormal in key names. If you really need a key name that ends with '{', don't use the JSON function, but just d['x{']:='y';

A minor downside may be that JSON calls with braces in the string constants, break the comment block if you want to comment out a section of code with braces, but there's always (* *) or //. It by far doesn't outweigh how elegant this new solution is, but it's a breaking change (existing code using '[' values won't work correctly any more), so instead of version 1.0.6 I think it deserves a jump in minor version number to 1.1.0.

While I was at it I fixed some minor issues with jsonV, so remember to update that one as well. Hope you enjoy the changes. (Next up may be replacing TMongoWire's bsonDoc with jsonDoc...)

Hoe het bij mij zit? Awel kijk. Het is moeilijk om te weten wat er gaat komen. Niemand kan de toekomst voorspellen. Soms is dat jammer, soms lastig. Heel misschien brengt het je angst, maar eigenlijk is het een uitdaging. We moeten ons behelpen met wat we weten van voorgane opvolgingen van oorzaak en gevolg. Dingen die je zelf hebt waargenomen, dingen die andere hebben waargenomen (en al dan niet willen delen). En daar moet je voor een stuk zelf naar op zoek. Je kan wachten tot je het nodige vanzelf oppikt, maar als je weet waar je naar moet zoeken vind je al snel heel veel meer. (Soms zelfs te veel en soms verkeerde dingen, maar da's een ander verhaal.)

Het moeilijkste in het leven is andere mensen dingen doen doen. Dat is zo de laatste paar jaar een motto van me aan het worden. Als het genoegen je te beurt zou vallen dat je van die kleine mensen fatsoen moet bij brengen zal je (nog beter) begrijpen wat ik bedoel. Maar het geldt in het algemeen. Zoals bij katten hoeden is de basis-regel dat je moet proberen maken dat ze willen doen wat het is dat jij wil dat ze doen, maar gelukkig is er nog voor wat hoort wat en misschien zelfs handel in goederen en diensten. Daarbij komt dan een sloot theorie over ethiek en afspraken en gemeenschapsgevoel enzovoort enzovoort...

Misschien lijken die twee niet verwant, maar op een vreemde manier is er wel een fijne samenwerken tussen die twee. Zo is het me gaan opvallen dat als mensen begrijpen dat je iets meer dan de doorsnee bezig bent met wat er komen zal, zowel op korte als lange termijn, en goed over de argumentatie hebt afgewogen, dan krijg je wel eens een doordachte vraag ten gronde over het een of het ander. En dan doet het deugd dat je kan antwoorden en al dat onderzoek van pas komt. Dat er misschien zelfs ook voor nog mensen naast jezelf iets duidelijk wordt over wat er komen gaat. Wat er nog moet gebeuren en wat niet. Dat schept een band. Je wisselt een beetje krediet en/of vertrouwen, en krijg je misschien al eens iets gedaan van mekaar.

Maar dat is een heel gevoelig evenwicht. Het doet goed als dat goed zit, maak ik me zorgen over als dat niet goed zit. Kortom, daar vul ik mijn 10 seconden mee tussen WM_ENDSESSION en Frequency Kenneth.

And for the people that do the speaking in English, a while ago I've written here around about the same things.

I've given it another try and, in part thanks to a Windows executable that performs well, and thanks to the gentle people over at the rethinkdb-dev group, I've got something working. It appears the basic functions you would expect from a driver work, but more advanced queries need some more extensive testing. (I've read about a driver testing harnass, but to replicate that around my driver, sounds like a separate project in its own.)

There are some strange sections in the design of ReSQL that show it was written for a weakly-typed language, but using variants and interfaces, I hope I've struck a nice balance between versatility and still having the best suitable methods showing up on auto-completion. Something that should be there may be hidden from view this way, but will probably be because of inaccuracy or error of my part.

Which leaves my in roughly the exact same position as I was closely after completing TMongoWire: having success learning about a new(ish) NoSQL DB the hard way by trying to connect to it directly with Delphi, but nothing in the way of an effective application that would use it for something vaguely useful. I've learned some people have effectively created things with TMongoWire (yey!) but I myself don't really have a good idea (or the motivation) to build something on its own that would use the features of these new database services to their best intent.

If you've got a good idea, please let me know. If you want to know more about TRethinkDB, let me know.

It is time to abolish the Union. Or incisively revise it, your call. A new leader is elected, and will soonish deliver the next State of the Union speach. It is not difficult to predict how it will start: "The State of the Union is strong." But in these challenged times, one may speculate it will contain some number of lies. The Union itself may not have the appearance of being in any other state than a strong one, primarily because partisan victory was upon more of the states than any prospector had foreseen. But the people are very much divided, and may be slow at reaping the benefit of hindsight.

The United States of America, as I see it, is three countries into one: An eastern seabord with longer history than the rest; a western seabord with younger infrastructure and a more liberal stance, but with a close eye on the environments, in part due to the earthquakes and drougths; and a less densily populated innermost region where the people are oriented around it's own set of drivers, such as agriculture and oil. The Union is above all a way to get these three to decently cooperate. If the Union fails, I would suggest it would be a good thing that the remnants of the Union would fall back upon these three regions.

In a new structure where the people elect the government of their region, these governments should put forward the individuals to make up a central organ, that only governs what the regions wish to share amongst them: the military, the monetary and a singular office for foreign relations.

But that's perhaps a simplistic view, that does not cater for the nuances and peculiarities required to organise a nation. Also, it is not my nation. Above all it's important there is a system that allows the people to have a say in how it is run.

Da's grappig. Ik deed laatst over de middag op het werk nog mijn uitleg over een ideetje waar ik mee speelde. Het gaat ook over twee munteenheden binnen een landseconomie, maar dan wel helemaal anders dan hier. Niekske protectionisme. Dus, als je even tijd hebt, je interesse is gewekt, en me het even wil laten uitleggen, hier komt het:

Stel, — de vraag is eigenlijk of een systeem zoals hiernavolgend uiteengezet hoegenaamd zou kunnen werken, en indien niet waar precies het op stuk loopt — we veranderen de landseconomie en de nodige wetgeving om te werken met twee munteenheden: de één is specifiek bedoeld voor handel in goederen, de ander is specifiek bedoeld (en gereserveerd) voor het vergoeden van werk. Ze zijn verhandelbaar aan een vlottende koers, en het omzetten naar elkaar gebeurt onder strikte regels. Doel van deze opstelling zou zijn om inflatie beter te kunnen beheersen, en de nood van een indexatieregeling weg te nemen. Ideaal is er geen inflatie op de munteenheid voor werk, of toch minder dan de andere. Werknemers en arbeiders krijgen dus een loon in werk-munt, maar wat betreft zelfstandigen, en in het algemeen handel in diensten ligt het mogelijk wat moeilijker. Momenteel, als je iets afneemt bij een zelfstandige is niet uitgeschreven precies welk deel van je vergoeding de aangekochte materialen dekt, en welk deel de gepresteerde tijd. Heel misschien wordt momenteel in bepaalde gevallen een soort fiscaal forfait gerekend om dit op te delen, maar als dit voorstel zou worden uitgewerkt, moet hier zeker duidelijkheid over komen. Om verder te gaan veronderstellen we voorlopig dat, mogelijk in samenwerking met verenigingen van zelfstandigen, per soort werk vooraf een aanvaardbare opdeling tussen de twee is afgesproken.

Een van de vele zaken die dan ook moet worden aangepast, zijn de maaltijdcheques. Als je ze ziet als een instrument van de staat om zeker een deeltje van de vergoeding van de werknemers sneller terug te zien in de vorm van BTW op voeding, dan kan je stellen dat zoals ze nu werken dit doel niet huldigen, al was het omdat ze nu 12 maanden geldig zijn. Gezien ze er ook voor zorgen dat een deel van de verloning zo niet meetelt voor persioenopbouw, geef je eigenlijk beter een rauwe belastingskorting. In het verlengde daarvan liggen ook andere denkoefeningen waar de staat zomaar aan elke burger een vast bedrag beschikbaar maakt. Of dit op grote schaal een basis-luiheid onder de burgerij zal losmaken, vernemen we mogelijk binnenkort naar aanleiding van praktijk-experimenten in Finland, Nederland en Zwitserland. In dit voorstel valt dat alles dus op de denkbeeldige breuklijn tussen de twee munteenheden. BTW hef je op de goederen-munt, en mogelijk niet meer op de werk-munt wat mogelijk de administratie sterk vereenvoudigt. Geld dat de staat normaal spendeert om de koopkracht te verbeteren, zou in dit model rechtstreeks worden aangewend om de koers te sturen tussen de twee munteenheden. Een vlottende koers staat rechtstreeks(er) in relatie met de tegenstrijdige belangen die hier spelen: door de werk-munt meer waard te maken in goederen-munt kan de koopkracht worden verhoogd, door de koers te drukken kan de staat waken over het draaien van de economische motor en zijn belastingsinkomsten verzekeren.

De Belg kennende zal er natuurlijk worden geprobeerd te foefelen. Je kan denken aan een belasting op omzettingen tussen de twee munten, maar dit is ten zeerste af te raden, want dan groeit er zeker een zwart circuit. Gezien voor de praktische uitwerking we op de (groot)banken aangewezen zijn, — in deze moderne tijden zou het toch moeten lukken om een extra munt volledig digitaal uit te bouwen? — zouden die waarschijnlijk ook graag iets afromen. Daar wordt best op aangeknoopt bij het lanceren van het systeem zodat het iedeen duidelijk wordt dat het loont om de aangeboden (officiële) kanalen te gebruiken. Dit kan door een structurele vergoeding uit te werken, die de kosten bij/van de bank(en) dekt, en aansluitingen bij boekhoudingen en fiscaliteit.

Misschien zijn er nog kansen om extra transparantie te bekomen, of administratieve vraagstukken zo op te lossen. Of heel misschien lijkt het wel een leuk idee, maar als je dit verder uitwerkt valt het ineen als een kaartenhuisje en is het in de praktijk volstrekt onwerkbaar. Wie weet, ik studeerde geen pol en sok. In elk geval staat het loodrecht op het voorstel van Le Pen, en gaan het niet tussen bedrijfen en het plebs, maar tussen werknemers en ondernemers gelijk, in de hoop voor iedereen voordeel en transparantie te bieden. En een extra bescherming tegen de grillen van de internationale economie.

Am I becoming an 'old hand'? Every time I see someone that doesn't have done as many years of work in Pascal, ask something about generics, I'm troubled again by this generics thing. I don't use it (much). I don't need it. I've been perfectly fine all of this work without them. Yes there is some hype nowadays about functional programming, but there's a grand glacial movement at the base of it: stong type systems. It all started even on the big metal, where some bright minds had the sense of taking writing logic into a domain of itself with a gentle nod to mathematics, statistics and other adjacent domains. And a lot of great work was done in this domain. A lot. Some great names that now stand on their own came forth.

A recurring theme throughout of it is type systems, or the lack thereof. Or having it bolted on at a later stage. So by the time K&R came along, they took what they needed, what they could make work for them, what they needed to summon the Unix spirit from the depths. Pascal existed at the time, but C, like JavaScript much later, was born out of necessity. Created by the people there and then that needed a new wrench to tighten the new bolts of a new machine. With personal selections here and there, and careful considerations that copy a bit of the zeitgeist of the time.

What's left for us to do is try not to repeat ourselves too much. One way is keep a look-out for patterns and in new projects, know when to apply which one. Or even better, when to switch to another one. This, I think, is where it's starting to get really tough for the younger generation. There's a lot of base-knowledge to cover, some of it is superceded or obsolete, but we can't replace it with studying patterns. I've seen newbies take it on as law, as if it's forbidden to deviate from the design even just a little. Attacking establishments on the base of their deviations from a pattern. There's something to say about attacking the establishment, but blind fury will get you nowhere.

So this generics thing... It also is in danger of being percieved as a rule, getting over-applied. As if the one true way is using it everywhere or as often as possible. Which is strange from where I stand, since I still see it as a convenience, where you can save on writing bits of code that would otherwise be painfully similar anyway. In operation the work would get handled by this extension to the type system, checking things for you in the background, making sure everything fits. But in the end the things that need to happen, occur much in the same way as if you would have written all those specialisations out in full.

That's why, as it turns out, I don't use them all that much. Perhaps it's also because of my fading belief in the object-oriented way. That too was presented some years ago as the way to go. Everything was supposed to be objects living on the internet. But it too kind of degraded in my view to a convenience, only one of the ways to manage the things inside of your program. Luckily before any of the ORM craze could tinge me. It makes you look around the horizon, in the past where you did more with arrays and pointers, and sideways where others pass messages around all the time. For now I feel like I'm stuck here, waiting for the ground to crack open and offer a new middle ground between interfaces, pointers, traits, channels, monads, and all of the above.

But that's just my humble opinion. I can't tell for sure what the next 60 years of programming machines will bring us.

Thing I learned today: 'method modifiers' are not reserved words. Apparently the Object Pascal parser 'consumes' these as labels that just happen to be there or not. Which means in other contexts they're free to use as identifiers, and so they're not reserved words... Strange but OK I guess as this keeps the tokenizer that does its job right before the parser, a little smaller... (How I stumbled upon this is a different story altogether, but one I may be sharing sometime soon...) I searched the documentation a bit and there they're called directives: Fundamental Syntactic Elements: Directives which I thought were things like {$XXX+}, but those are compiler directives...

So these variable names are accepted by the parser/compiler (it doesn't mind the smell at all):

I just noticed these went missing since I converted this website to a 'static blog' kind of thing. So I delved up the old backups and fetched these back from the old mdb, and put them up here.

First a bit about me, a little CV:

I started at a young age with a little BASIC on a CP/M (on a Kaypro II!),

later GW-BASIC but soon enough Turbo Pascal versions 3, 5 and 7. Almost by chance, as my father bought Turbo Pascal and Turbo C++ on a whim without really knowing what it was. My brother who is 4 years older also did some Pascal, but picked-up on C++ and later did heavy work in assembler as well.

Later, when I studied for what is now equivalent to Bachelor Computer Science, I was introduced to Delphi 3. Again almost by chance as someone I showed old Pascal DOS programs to, told me there was something new for Windows. After that I moved to Delphi versions 5 and then 7.

By then I attained my degree and started with a web-development firm. I first did PHP and later Cold Fusion, but did a lot of work of my own on Delphi-projects as a hobby.

But as hobbies go I did not put any money in new Delphi versions. Also I did not know any other Delphi developers professionally. I later heard about the schism in Delphi-world as the first versions after Delphi 7 had put off a lot of people because the IDE and editor was less good.

So I'm afraid I'm also one of those people that sticks with Delphi 7, but in part by consequence and only in part by choice. Still, thanks to the careful design of the Object Pascal language, code written for/with Delphi 7 still works really well in modern Delphi versions (just with taking care about a few little things like using AnsiString/WideString where possible.)

A funny fact is that I only learned about SQL and databases later on, around when I started to work on websites. I have done a lot of work in Delphi and not used the data-aware components! Also by doing PHP, ASP and Cold Fusion and developing websites without an IDE (just a source editor), I really disliked the different options you had to develop websites in Delphi. I'm conviced the IDE/RAD component eco-sphere is toxic to the performance-centric build-up of a website and will always make important choices for you that will cause problems later on when the project gets bigger.

So I started xxm. The Delphi compiler is so fast, it is possible to have a DLL that runs a website, when source-files change and you hit Refresh, to unload the DLL, compile a new one, load it and have it serve the page with updated logic. This way, developing a website without IDE or components would be almost the same as PHP and Cold Fusion, except in Object Pascal, and the website would perform as fast as a native service.

Programming a website like this reduces the work to creating a stream of HTML, and any visual components you would use offer extra overhead which takes a bit of your performance away. So that's where my search for 'really thin wrappers' comes from. When using Microsoft Access MDB's at first, and later MS SQL Server, Microsoft already developed the ADO components to do the heavy lifting, and thanks to the integration of ActiveX interfaces (IUnknown) and the type library importer (to generate ADODB_TLB.pas and others like MSXML2_TLB.pas) it's easy to use the ADO components directly (mainly Connection, Command and RecordSet, see xxmData.pas).

Later I also checked out other databases, but was displeased with the performance of the ODBC connector to SQLite databases. Other existing Delphi SQLite connectors were for modern Delphi versions and had long lists of 'overload' properties, which I really dislike. Using ADO much of the data-transport is based on OleVariant and I like that. Then I discovered the C API to SQLite is relatively straight-forward, and created TSQLite. Also PostgreSQL and MySQL/MariaDB have a C API DLL ready to use to connect almost directly. Since their SQL dialect is very similar, it should be easy to switch between the two (e.g. start a project off on SQLite, and later switch to PostgreSQL to serve a growing user base) so I created DataLank to take a lot of the work of switching away.

After that I also wanted to check out some of the new 'NoSQL' databases. I had no experience, and thought if I can write a connector I would understand more about how these databases worked.

With Redis I learned it had a human readable protocol, but was very basic and very much removed from the SQL world.

With CouchDB I learned some 'NoSQL' databases concentrate on storing (JSON) documents and can perhaps do more server-side but can do it fully separate from the front-end.

I tried RethinkDB several times: first I learned about Protocol Buffers and made Delphi support myself because I couldn't find any. By the time I got that working (remember, I'm a hobbyist programmer that only puts a few hours a week in my own projects), RethinkDB was moving away from 'ql2.proto' but was not ready with their new binary-level protocol. And the next time I checked, they were in some kind of trouble.

With MongoDB, I think a nice balance exists between storage of documents, and other features you need to do specific manipulations or queries. I was lucky to find my way through the layers of documentation to what I needed to know about the binary protcol. First BSON and then the wire protocol, both show careful design and are not needlessly complicated.

There are a few more on my to-do-list, but I kind of lost interest. I'm using TSQLite the most for the moment, and will probably start on a new connector if I'm sure there's the drive for it from a new project that's bound to have to run on this or that newfangled database. One big change I consider making somewhat soon, is reworking everything BSON but based on my own JSON parser I created because I needed it for a project that otherwise didn't need the bits specific for BSON...

v1.2.3.497
- fixed issue about division by zero at start-up
- now open source on github