On 11/8/10 2:07 PM, oldtimer wrote:
Mon, 08 Nov 2010 13:06:23 -0800, Andrei Alexandrescu wrote:
On 11/8/10 12:48 PM, bearophile wrote:
Andrei:
A longer term perception problem is that some might think the
_design_ of such features has unsolvable issues,
That's my fear (so it's problem, even if it's just a perceived
imaginary problem). But I am not in a rush, so I am not going
away.
I know of no major design flaws,
Uhm... For example the whole implementation of nonnullables in D2
is already becoming a big design flaw.
I think we have different definitions of what a design flaw is.
C++ export was a design flaw. PHP also has several design flaws.
The fact that you can't implement non-nullable references in e.g.
C++ or Java is not a design flaw in those respective languages.
Generally I think the case of non-null references is overstated.
It doesn't deserve that much agitation. It's being mentioned as a
life or death matter in spite of the simple evidence that many
mainstream languages are doing well without it, and without a
landslide popular vote for adding it. (And it's not a flaw in their
design that they don't make non-null references implementable.)
For some members of the community D has been a learning process; D
has introduced the members to new features in language design. It has
been an informal trial and error process. Even Walter has learned a
lot, which is nice. This has caused at least two problems. The
requirements have changed and developers expect now more advanced
functionality. The amount of suggested design ideas has grown, but
the development speed hasn't scaled as well. The development process
doesn't scale up since almost no documentation gets written. The
expectations may be unrealistically high. Sometimes the expectations
are realistic, but the combination of legacy design choices in D1 and
D2, TDPL pressure, D2 stabilization pressure, and D2 schedule has
brought a suboptimal implementation.
The second problem is, the developers now expect more knowledge from
the language designer. It gets really frustating if we need to start
teaching him to get better features. This is universal - if your kids
go to a really bad public school, you might want to start teaching
them at home or consider some private school (= other language). If
your lawmakers don't understand the laws they pass, you might
consider moving to another country (= other language/platform). I can
say I'm not really convinced. I'm probably not the only one.
It's bad for the publicity that digitalmars can't articulate any kind
of official *written* roadmap for new features. You don't want to
dedicate much (financial) effort on projects using the cowboy coding
paradigm. The project has a trac page. Writing down: "1. 64-bit dmd
(N months) 2. concurrency fixes (M months) etc." shouldn't take too
long. If my guess is right, you can even copy'n'paste it straight
away. But you also need to update it! This is basic project
management these days.
All very good points. Walter is generally shy to put forth agenda that
are difficult to estimate and meet. There have been, for example,
unexpected (and unexpectedly long) setbacks of the 64-bit release due to
seemingly minor hitches (e.g. C-style variadics). At any rate, I, too,
think a sort of general agenda would be useful to keep updated.
One simple thing that I've recently done (my free time has been
relatively scarce lately) was to work on bug reports starting from the
oldest first. (I solved a bunch, too - it took me a couple of days to go
through like 70.) The oldest non-enhancement non-lowpri unsolved bug is
a metric that people will look at, and limiting the age of the oldest
bug gives people a sense that all reports will be looked at in a fair
manner.
One reason why Python/Ruby are so popular is their terse syntax. The
library based solutions in D2 aren't that cute. Some developers value
the beauty of code. If the solutions are too ugly, people will switch
to other languages. Compare Pascal with C, the languages are quite
similar modulo syntactic changes. The other one won.
I agree in spirit, but beyond some point this all becomes a manner of
taste. And library solutions scale in many ways much better than
solutions built into the language - I mean look at Python's "print". I'd
actually prefer NonNull!T over @T or @nonnull T etc. What I prefer even
more about NonNull!T is that it uses general language facilities that
can be used for NonZero!T or Bounded!T etc. as discussed.
For commercial users of D the feature set matters a lot. If D2 can't
deliver the power of D1& Tango, there's nothing you can do.
What I can do is work on D and encourage others to do the same.
A sane
corporation would not donate code to Phobos just because its quality
is worse and it's an empathetic thing to do. Corporations don't work
this way. Those who are fixing Phobos are individual zealots swimming
against the stream. They have a very long term technical agenda (D2
will beat comtemporary technologies in 10+ years), a religious
agenda, nothing better to do, or some kind of mental illness.
Corporations aren't interested in very long term technical issues,
they care about short term solutions (1 to 36 months).
Some of us hoped D2 would solve the ugly corner cases of the language
and unify a lot of features. Unfortunately not any kind of macro
system was implemented, templates got even uglier, the alias this
hack, snake oil template!X syntax shortcut and all kinds of
unnecessary features were added. Of course also good features were
added, but overall this just isn't good enough.
I was expecting a number of responses from under the cloak of anonymity
(i.e. from people whose identity would immediately reveal their agenda),
and as far as those go this post was surprisingly good up until the last
paragraphs. Here is unfortunately where the tone of the discourse degrades.
Let's not forget that a lot of good work - including Tango itself - has
been done by "zealots". It also seems that disingenuously attempting to
frame D2 contributors as weirdos with a religious agenda, nothing better
to do, or mentally ill is a display of an emotional position in the
matter that instantly reduces the quality of any point made and that of
the dialog in general. This all being told from under anonymity doesn't
help either.
I understand you would have liked macros, dislike the paren-less
shortcut notation A!B instead of A!(B), and dislike some untold features
that you find as unnecessary. It would be great if you substantiated
these points while leaving the subjective outside. For example, I am
confused by the shortcut A!B being snake oil or alias this being a hack.
The former is a mere convenience and the latter is a well-motivated feature.
Andrei
