Thread: Why nothing but beginner questions?

Originally Posted by Nominal Animal

In any case, I still believe there are people working on solving actual interesting problems using C (or C++ or Fortran); I'm just bewildered as to where the heck they're discussing the issues at.

Tool-specific development-oriented (rather than end-user support type) mailing lists come to mind.

Originally Posted by Nominal Animal

In the scientific literature, both math and computing papers have the same extremely annoying feature, in my opinion: the most interesting bit, the development of a solution, is often skipped by using circuitous verbiage like "find a set S that fulfills ...", "having a tree T such that ..."; the actual interesting part, programming-wise, being how such a set is found efficiently, or how to construct such a tree efficiently, and so on. Basically, the papers either tell you there is such a way, or that there is an optimal way with specific characteristics, and spend a lot of effort in proving it, but they typically omit the actual interesting bits.

The "interesting bits" are probably, more often than not, proprietary trade secrets.

Originally Posted by Nominal Animal

Circling back to my original question, this does mean that a lot of interesting low-level library work must happen in C (or in C++ or Objective-C). This includes stuff like drivers, high-performance libraries, and so on.

So, where the heck are those developers discussing their development features?

I kind of alluded to this earlier... the explanation is that we are very, very busy. And so conversation is typically limited to only what is required to get past the current set of issues. These discussions happen over email, phone, sometimes at conferences face-to-face. I've flown on airplanes to have these conversations. Sometimes there are private mailing lists.

In the scientific literature, both math and computing papers have the same extremely annoying feature, in my opinion: the most interesting bit, the development of a solution, is often skipped by using circuitous verbiage like "find a set S that fulfills ...", "having a tree T such that ..."; the actual interesting part, programming-wise, being how such a set is found efficiently, or how to construct such a tree efficiently, and so on. Basically, the papers either tell you there is such a way, or that there is an optimal way with specific characteristics, and spend a lot of effort in proving it, but they typically omit the actual interesting bits.

Yup. It bugs me too, or at least it did, before I managed to get into "the circle." It's absurd but there are social tiers and people just don't want to waste their time communicating with the outer tiers.

Case in point, look at my posting history the last couple months. Basically nil. I've been very busy. I'd love to talk about crazy-specialized topics, but in my experience people will attempt to argue about things that frankly, they have no business expressing their opinion on. Rather than being an ***hole I just refrain from the conversation most of the time. It's kind of like doctors or lawyers. Incredibly close-lipped. You try to spread knowledge, and in come the horde of morons and people trying to manipulate and take advantage of you, or try to hold you accountable for things you just don't want to be accountable for. Thus the silence.

Originally Posted by brewbuck

You try to spread knowledge, and in come the horde of morons and people trying to manipulate and take advantage of you, or try to hold you accountable for things you just don't want to be accountable for. Thus the silence.

Life is sometimes like getting beaten to death by feathers. Each blow is inconsequential on it's own, but the effect can accumulate until you want to detach yourself.

Originally Posted by stevesmithx

───────────────███───────────────
─────────────██▀─▀██─────────────
───────────██▀─────▀██───────────
─────────██▀──▄▄▄▄▄──▀██─────────
───────██▀──▄▀─────▀▄──▀██───────
─────██▀──▄▀───███───▀▄──▀██─────
───██▀────▀▄───▀▀▀───▄▀────▀██───
─██▀────────▀▄▄▄▄▄▄▄▀────────▀██─
█▀─────────────────────────────▀█
█████████████████████████████████
─────────────────────────────────

That time the Illuminati got tired of putting subliminal frames in music videos, and decided to dominate ASCII art instead.

I've stopped seeing Illuminati. Now all I see is Yung Venuz.

Originally Posted by Nominal Animal

To be honest, it does show on the end results.

No, I absolutely do not mean yours personally/specifically. I mean, the code quality of scientific software at least is pretty darn poor. Buggy is the norm. Feed one unexpected data, outside the small range the original programmer thought about (of course undocumented anywhere), or more than its programmer needed to expect, and they're more likely to crash than work right. Or worse, silently produce bogus results.

It's a valid criticism. But obviously it's a huge range. I know we've disagreed with each other on this in the past, but I truly believe that wider adoption of C++ in embedded and firmware coding is going to improve things if combined with proper development and testing methods. I've got a younger engineer I've been working with for a year or two now and he is pushing me on a daily basis to do things better, and I'm benefiting from it. This is a person who fully unit tests every single thing he does, even things that most other people think of as "just scripts." As this practice spreads I think things will improve and move software engineering more in the direction of an actual engineering discipline.

From a certain perspective it might look like C and C++ are "dying" but from my viewpoint that isn't what's going on at all.

It is unfortunate that most research teams do not have "problem-solving tech janitor/computing cluster admin" as part of their teams. I'd fit that position pretty darn well, I think.

That's a position almost any company can use, and yet one of the hardest things to justify to management. Good engineering for its own sake is a really hard thing to justify in a capitalist economy. Revenues are always going to be more important.

Originally Posted by brewbuck

I've got a younger engineer I've been working with for a year or two now and he is pushing me on a daily basis to do things better, and I'm benefiting from it. This is a person who fully unit tests every single thing he does, even things that most other people think of as "just scripts."

We need your guy. One of the troubles we have been facing lately is our ability to integrate TDD into our largest projects. The need for a disciplined and well structured modular method for efficient TDD is driving us up the wall.

TDD can be applied regardless of the project size. But I think from my experience that it starts to break apart on small teams. You just don't have the enough manpower to ensure there are few blind spots or mistakes in your tests. Or that you write enough integration tests. You end up relying all your testing on the very people that write the code. Maybe there is some magic he could concoct. The biggest problem i'm having with our attempts at introducing agile-like methods in our projects is the fact they actually demand an increase in the number of software developers. And we can't have that.

Originally Posted by Epy

If you're getting bored you can check out the bugtrackers for OCE or gfortran

I swear, if we ever have a thread about the difficulties of being a single parent, I will be disappointed if you don't mention Fortran somewhere.

Originally Posted by Mario F.

The biggest problem i'm having with our attempts at introducing agile-like methods in our projects is the fact they actually demand an increase in the number of software developers. And we can't have that.

Do you really think you need more developers? Or can you just accept the reduced velocity?

The thing with TDD is it feels like you're going slower, but you end up going faster by spending less time wallowing in bugfix mode.

I know the difficulty of really making it work, though.

Bring back CommonTater!

Originally Posted by camel-man

Bring back CommonTater!

And Grumpy, while we're at it. I miss Grumpy.