[dmd-internals] Oldest five bugs

[David Simcha]

On Wed, Jan 18, 2012 at 5:24 PM, Don Clugston <dclugston at googlemail.com>wrote:

>
> More importantly, I think this attention is misplaced.
> Organizationally, the attention needs to be on Phobos, which, in stark
> contrast to DMD, has not noticeably benefited from the move to git.
> Look at the changelogs. There are zillions of compiler fixes, and
> practically nothing for Phobos.
> Phobos is not moving. What can we do to improve the situation?
>

One of the biggest things holding back Phobos development is that so many
of the things that are missing are high-level design issues as much as
implementation issues.  (For examples see my previous post that touched on
this:
http://lists.puremagic.com/pipermail/digitalmars-d/2011-November/115320.html).


Implementation and low-level design are a lot more parallelizable than
high-level design because they can be done on the spur of the moment
whenever one has a free hour or two.  High-level design tends to require
long discussions if done collaboratively and/or long blocks of free time to
hold a lot of stuff in one's head if done alone.  Furthermore, if someone
just hacks something together and the community doesn't like the high-level
design, the work will be completely wasted.  This is a deterrent.

This is compounded by the newsgroup not being the best medium to discuss
design.  Asynchronous communication is good for quick messages.  In-depth
discussion requires that the people discussing hold a lot of information in
their head at once.  When a discussion gets too deep on the newsgroup, it
becomes hard to follow because the amount of context that has to be
reloaded into one's brain to make sense of the latest post becomes too
high.  Maybe the core Phobos devs should occasionally use some more
synchronous form of communication (e.g. Skype meetings, chat, etc.) to
hammer out high-level design issues.

Lastly, design issues tend to require intimate familiarity with use cases.
I would have no problem implementing a container library (it's just basic
data structures) but I don't feel qualified to do the high-level design
here because I don't understand all the use cases that make it a
non-trivial problem.  Probably very few people have diverse enough
experience to have a deep understanding of the tradeoffs involved.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/dmd-internals/attachments/20120118/3e40881e/attachment.html>