[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Architecture qualification

On 12861 March 1977, Steve McIntyre wrote:

>>There's a related question, which I just realised wasn't actually
>>explicit - does it make sense to add an architecture to testing at this
>>stage of the process which we don't think is releasable?  My memory of
>>previous discussions is that the general answer was "no", although this
>>possibly depends on how one views the purpose of the testing suite.
> Definitely not, IMHO.

> How hard are the RT / ftpteam going to stick to the "ship with Wheezy
> or you're out" agreement as written in http://wiki.debian.org/Debian_GNU/Hurd ?
> Is Hurd at the point where we could *reasonably* ship it as (at least) a
> technology preview?

I am pretty set on that. There isn't much point in coming to an
agreement just to kick that out the door only because the time has
come to actually do something about it. (And why didn't inclusion of
hurd got a topic like, half a year or more ago?)

There is only one thing I would agree on: If the RT decides to not
include them in wheezy but add them to wheezy+1 right after wheezy is
released (so we would be doing it during the process) and keep them
there for the next release, then fine. Thats not exactly what we agreed
on, but would be a workable compromise.

> I'm unconvinced that it is, being brutally honest.


I also think it is way to late before the freeze to add something like a
whole architecture (and hurd is more than just a plain architecture)
*now*, where we already kick maintainers for uncoordinated package
transitions, prepare to kick people putting (uncoordinated) SONAME
changes into NEW and generally want to have a freeze RSN.

Also, what is really changed when we do this?

- hurd is no longer on the main mirrors. But "only" on those carrying
  * So what? Yay, we suddenly stop mirroring something to thousands of
    places which is used by less than anything else.  Don't tell me
    there are so many users of hurd that it really warrants the wide spread
    mirroring. Especially with it being very limited on desktops and
    probably serious servers[fn:1] too? (Reading from
    http://wiki.debian.org/Debian_GNU/Hurd and links)
- hurd can come back into the main archive following the usual archive
  qualification every other new addition has to follow. Clean, simple,
  straight forward.
- We are rid of a special case of an unstable/experimental only
  architecture that often keeps pretty outdated packages in the archive.
  Less general maintenance work.

Of course the above is my personal opinion and ftpmaster is a team. It
might happen the rest disagrees with me and tells me to sod off, but I
think thats highly unlikely, from what I know.


[fn:1] That is, not just "yeah, here, look, there is a server", but
"there is a server that actually has a production used application on
it. Is HA and whatnot, and $company does $lotsathingswithit, relies on

bye, Joerg
Ich will ein anderes Telefon, das hier klingelt immer!

Reply to: