Xorg 1.5 missed the train?

[Suren Karapetyan]

Les Mikesell wrote:
> Suren Karapetyan wrote:
>>
>>> I assume that was an attempt at humor....  But, it makes it hard to 
>>> claim that you didn't have some inside information about when the 
>>> interface was going to stop changing.  In another company that sort 
>>> of thing might be called anti-competitive behavior.
>>>
>>
>> Guys let's stop using the argument "they didn't know it was stable"...
>> If You're writing a driver for Your product and not just an ordinary 
>> userspace thing, but a driver half of which sits in the kernel and the 
>> other half in X, You'll HAVE TO have a guy (or maybe many more) who 
>> will be doing just that and nothing else.
> 
> Yes...  But this may not be the guy that decides when an officially 
> supported driver is announced and released.

Yep... You're right. And the guy who decides will most likely also say: 
"How many people use that X *thing*? Less then 1000? There is no way 
I'll pay for writing the driver for it."

> 
>> And I bet if someone's job is writing an Xorg driver, he would at 
>> least be signed to the -devel mailing list and would checkout from 
>> CVS/SVN/GIT/... at least once a week to watch where the development is.
> 
> Yes, so if someone mentioned that it was maybe, probably stable a week 
> ago without being prepared to call it a release, you might expect said 
> programmer to have noticed by now, but it hardly seems fair to expect 
> him or his company to commit to a release at that point either.

It may not be true for his company, but it does make sense for him.
The point is if the ABI seems to be stable but changes after a week, I 
wouldn't expect it to change much.
And if I was THE programmer and I knew that after a month or so I'll be 
ordered to write a driver for the new version of Xorg, I would start 
thinking (read: compile-debug, compile-debug,...) about it as soon as 
the ABI *seemed* to be stable.

> 
>> And don't tell that's not the case with Windows. Of course it isn't... 
>> But we aren't talking about a windows programmer who is writing Xorg 
>> driver as a hobby in the first time in his life and doesn't know that 
>> ABI's aren't very loved in FOSS world. We are speaking about a *nix 
>> programmer.
> 
> *nix doesn't have much to do with refusing to standardize interfaces, 
> that's exclusively Linus's territory.  I think we'll see something 
> different when Red Hat does their release.
> 

I wasn't talking about *nix. I was talking about FOSS.
And it isn't about "standardize interfaces", it's about stable ABI.
Enterprise software vendors have the problem of having to support old 
versions of their software, even if they don't want. That's the reason 
of the old LM password hashes from Windows 95 till Vista, and the old 
LinuxThreads compatibility library in RHEL. This isn't true for FOSS 
developers. They don't have to be a "hostage" of the software they write.
Nothing limits the speed of changes in FOSS software.
That's why if a FOSS project sees even a small benefit from breaking the 
ABI, it won't usually think twice.