ETAs for libusbx operability and first release?

[Pete Batard]

On 2012.01.30 11:20, Segher Boessenkool wrote:
> I'm back,

Welcome back.

If this absence was planned, and considering that we are still in a 
stage where we can't progress on the fork unless you're there, an ETA of 
when you would be back would have been nice (see below)...

> and I'll get the infrastructure sorted ASAP.

Great. If you need help one way or another, shout away.

> The more important part, the code, is non-existent (no one has
> posted any patches yet!)

Don't know about everybody else, but personally I'm waiting for git to 
be up on the server so that we can bring the current existing libusb 
repos and then apply submit and apply patches.

> What are we basing on, the current state of libusb.git master?

I'd say so. It just needs a couple of patches to be applied to be in 
releasable state IMO.

> Please start sending patches. Or discuss what we should
> include now (not too much please!)

OK, I'll do that for the must have patches (which are the ones that were 
posted by Michael on libusb).

But without a public git repo they can be applied to, it sounds a bit 
abstract and I'd like to avoid a repeat of what currently happens on 
libusb, with the same patch is submitted 3 or 4 times on the mailing 
list, where everybody agrees it should be applied, but where it never 
actually ends up in git.

If we're going to discuss patches, having the possibility to apply the 
ones we think are good to go immediately is a must have.

>> What's more, since I'd like to avoid falling into the pitfalls of
>> "it's ready when it's ready",
>
> You have a problem then. It will not be ready before it is
> ready! Complaining about it doesn't change a thing.

What I'm complaining about here is people in charge of performing a 
*blocking* task, such as setting up the server, going away for weeks 
without any form of prior announcement, and leaving everybody else 
wonder what's going on...

I don't have an issue with you going away for as long as you want, but 
it'd be nicer if you announced it beforehand since there's currently a 
bunch of people depending on the completion of this task (of course, I 
don't know your circumstances, so maybe you couldn't avoid it).

Once the server is operational, that won't be much of a concern, but to 
me, this 2 weeks black hole where people were left to wonder why a task 
that seemed to be on track got halted without notice is a stark reminder 
of one of the main reason we forked from libusb in the first place...

> I can give you a plan though. Plan:
>
> 1) Put whatever we want in the first libusbx release in a tree;
> 2) ???;
> 3) Release.
>
> I estimate this to take two weeks.

Yay! Our first ETA! (I don't think the plan was ever in question)

Now at least the rest of us have an idea of where we are headed, and can 
set our tasks and availability accordingly (and we also have an idea of 
when you think you'll be able to complete the setting up the server).

ETAs are movable targets of course, but at least we know where we're going.

>> which we know all too well the damage of from libusb,
>
> First, that is not the problem there.

See above.

I'll make no mystery that I am personally more than fed up with people 
having the means of blocking a project and actually acting as if they 
can block it without providing a reason. IMO, this shows a lack of 
consideration for anybody downstream.

If people are depending on you, the least you can do is let them know 
why a task is halted, and provide them an ETA of when you think it can 
be resumed. That's how the healthy projects I know seem to do it.

> Second, you cannot compare
> this first fork release to a regular release anyway.

Agreed. But I think the plan is for the first release of libusbx to do 
what should have long be done for libusb 1.0.9, which is:

- apply the last remaining critical patches that are needed for a 
release (I think they are all Windows related at this stage, but if 
anybody sees non Windows related ones that should go in, let us know)
- leave anything that can wait for 1.0.10 or later (or you can try your 
luck with small patches, but if everybody does it and it's going to 
delay 1.0.9, I'd motion to reject them all)
- release the first libusbx-1.0.9

> "Two weeks".

That's all I want to know. And the current ETA also sounds reasonable to 
me, provided we don't hit a snag with the infrastructure.

Regards,

/Pete