March 14, 2005
>> PS. Then again, Thomas already wrote a patch to remove usage
>>     of printf from the library but I haven't seen it applied ?
>>     (change "printf" into "writef", that was, in Phobos tests)
>
>Yes, my point is that nothing will be done unless Walter wants it done or has the time to get it done.  Submitted patches are a good idea only if they are guaranteed to make it into the update .  Otherwise they're a waste of time (I think).  At present there's no guarantee any contributions will make it in.  That's the way things work with D.  This system provides us with little motivation for participation in language and library improvement.

I agree to a point. When I've sent Walter things for phobos he has responded and we started an exchange about the proposed change. Some went in and some didn't. I disagree that all patches should be guaranteed to make it into the update. That would result in chaos. Submitting "arbitrary" patches is a waste of time. If we knew more about what kinds of patches Walter will accept and how to comminucate these changes to him it probably would make for more people submitting patches. For example, Walter could say "all patches must be posted to the newsgroup for discussion and in the email to me send the link to the discussion so I can see how the community feels about it". Or "all patches must have a backwards-compatiblity impact statement". I also think Walter doesn't like patch files - he prefers entire modules that he can just drop in. I can imagine he would ignore stuff like "just change foo in blah.d and see what doesn't compile and change those calls to bar". Anyhow, I think there's plenty more we can be doing to help Walter's through-put.


March 15, 2005
Ben Hinkle wrote:
>>>PS. Then again, Thomas already wrote a patch to remove usage
>>>    of printf from the library but I haven't seen it applied ?
>>>    (change "printf" into "writef", that was, in Phobos tests)
>>
>>Yes, my point is that nothing will be done unless Walter wants it done or has the time to get it done.  Submitted patches are a good idea only if they are guaranteed to make it into the update .  Otherwise they're a waste of time (I think).  At present there's no guarantee any contributions will make it in.  That's the way things work with D.  This system provides us with little motivation for participation in language and library improvement.
> 
> 
> I agree to a point. When I've sent Walter things for phobos he has responded and
> we started an exchange about the proposed change. Some went in and some didn't.
> I disagree that all patches should be guaranteed to make it into the update.
> That would result in chaos. Submitting "arbitrary" patches is a waste of time.
> If we knew more about what kinds of patches Walter will accept and how to
> comminucate these changes to him it probably would make for more people
> submitting patches. For example, Walter could say "all patches must be posted to
> the newsgroup for discussion and in the email to me send the link to the
> discussion so I can see how the community feels about it". Or "all patches must
> have a backwards-compatiblity impact statement". I also think Walter doesn't
> like patch files - he prefers entire modules that he can just drop in. I can
> imagine he would ignore stuff like "just change foo in blah.d and see what
> doesn't compile and change those calls to bar". Anyhow, I think there's plenty
> more we can be doing to help Walter's through-put.
> 
> 

Yes, you are correct in that regard. It wasn't my intention to suggest that he arbitrarily accept all patches.  I'm saying there is a problem with communication and organization, as ofted noted in the past.  After all these years, a system is still not in place to ease the patch submission process. As it currently stands, it's a "wait and see" thing with Walter.  That's got to dampen the motivation of even the toughest and most dedicated patch submitters... except for Anders, of course. ;-)

And the problem will continue and continue.  We'll go through that same old endless cycle of frustrated members asking why there's no bugzilla, why we can't help out if the language trully is open, why we have to wait for version releases to fix bugs, etc.  We'll continue to lose quality members as fast as we gain looky lu's and trollers.  We'll gain more people who are eager to offer D superficial praise after light evalution, and lose the critical thinkers that examine, stress, and critique the compiler into perfection.

Walter has a huge load with tons of backlog of fixes to be done. Between answering newsgroup posts, coding dmd fixes, improving his C++ compiler, and living his everyday life, I don't know how he manages.  I applaud him and thank him for all he's done so far.  D is certainly experiencing a surge in popularity and recognition; but I think something drastic has to be done with the system before this huge bubble bursts.  Suggestions have been made in the past only to fall on deaf ears.  Division of labor, delegation of tasks, and moderated discussion (ie open/closed committee) seem to be the way to go, if not now, at least eventually as the steam roll effect continues (post 1.0 perhaps; that 1.0 can be a stake in the ground for more than one purpose).  I strongly doubt the current system is going to continue to work.

-JJR
March 15, 2005
John Reimer wrote:

> Yes, you are correct in that regard. It wasn't my intention to suggest that he arbitrarily accept all patches.  I'm saying there is a problem with communication and organization, as ofted noted in the past.  After all these years, a system is still not in place to ease the patch submission process.

This is true. Other projects that have much more activity are faster
in accepting (or rejecting) patches, because they have a much better
infra-structure set up with systems like bugzilla or cvs (/subversion)

For instance, all the Doxygen changes made for D are already online ?

> As it currently stands, it's a "wait and see" thing with Walter.  That's got to dampen the motivation of even the toughest and most dedicated patch submitters... except for Anders, of course. ;-)

I'm using a package manager, which let's me easily test out new patches
and also downgrade back to old releases should they not be working out.
Using patches also makes it easier to upgrade versions from upstream.

But of course, I am much more likely to work on GDC than on DMD...

--anders
1 2
Next ›   Last »