Lubomir Sedlacik wrote:
> On Sat, Jul 21, 2007 at 06:40:59PM +0000, Jonathan Perkin wrote:
>> * On 2007-07-21 at 18:34 UTC, Jan Schaumann wrote:
>>> The platform not being labeled as "supported" will not have any
>>> effect on the packages continuing to build if they did and failing
>>> to build if they didn't.
>>>
>>> I see the label "supported" as meaning "people are actually working
>>> on these on a regular basis".  If your platform is not listed here,
>>> then you still may have a fantastic experience using pkgsrc.
>> Therefore, in light of this, I have to ask the question - what on
>> earth is the point of this "supported" label, if at the end of the day
>> it's ultimately a pointless and meaningless one at best, and confusing
>> and misleading at worst (when a "supported" platform actually builds
>> less packages than one which isn't)?
>>

The point is you don't have to find out for yourself whether it will
work or not. It's been tried and either failed or worked.

>> Where's the benefit?
>  Where's the net gain for pkgsrc?

. Shared knowledge of experience.
. Pkgsrc will be easier to adopt (sell).
. Potentially more users of pkgsrc.
. The task of supporting a platform is defined and can now
  be assigned, budgeted, paid for etc.


> what purpose would this information serve?  what practical problem are
> you trying to solve?  i certainly don't see the point.

From the perspective of someone who would like to use pkgsrc in
production environments:

Support is something that is heavily considered when selecting a
platform or system. Systems administrators who want to use pkgsrc will
be able to "sell" it to their superiors as a supported product.

If "supported" just means that someone tried it got the bootstrap to
work in the past, it is not good enough for an admin or even for a
consultant that wants to provide support. Not to mention telling your
boss you want to use an un-tried system to build packages.

What Jan is saying (I think) is that if someone is doing bulk builds,
then there is immediately enough awareness of what works and what
doesn't. If bulk builds are being done, then we already know bootstrap
works and that there is a compiler that is usable on the particular
platform. And this on a regular basis (quarterly).

Maybe there needs to be a threshold of certain mandatory packages being
built for a platform to be considered supported. But there should be
something measurable.

If nobody is doing bulk builds, it's a "good luck" situation.

If there is no measurable way of saying what is supported, then nothing
is supported.

Another benefit is that if someone really wants pkgsrc on a certain
platform they could be compelled to PAY someone to do what needs to be
done to make the platform supported. That's a measurable task that can
be assigned, budgeted etc.

> another thing is that i can pretty much guarantee that the list will
> bitrot within a year because nobody will keep it in sync with reality.
> heck, even such a simple thing as the name of the currently supported
> stable branch is often forgotten to be updated on the relevant web site!

That's a different problem.

There needs to be QA for the support. Regular checks need to be done on
each branch to see what can be considered supported. Make sure the
responsible persons are in fact actively involved.

This should not be difficult to maintain.

> i can point you to a lot of other things which need improving with the
> maintenance of the stable branches and the binary packages archive on
> ftp.NetBSD.org instead of making up arbitrary criteria to call pkgsrc
> on a given platform this or that.
> 

How is it arbitrary? The presence of bulk builds is a measurable,
quantifiable, way of determining that people are actively working on a
platform.

I think a larger discussion over "how to use pkgsrc in a production
environment" may be useful. What is "supported"? What can be expected?

These things need to be defined.

If pkgsrc wants to be taken seriously then it should certainly step up
to the plate and take responsibility for the support of platforms that
are being worked on.

If we tell everyone "use at your own risk" then pkgsrc will always be an
experimental system. Is that what we want?

Louis