Re: [gentoo-dev] Please use land-misc herd where appropriate! No no-herd madness!!!

[George Shapovalov]

Ok, to ease this off I should state up front that, while I strongly feel that 
we should adhere to existing regulations and common sense as much as 
possible, practical considerations (see end of this message) do not allow to 
just outright kill no-herd entries which we accumulated plenty already. Thus 
please observe the wording I used in my original message:
> > This is a big fat reminder that no-herd is really against the policy and
> > should not be used
> > *if you can find appropriate herd*!!
(see the emphasis)


However since this is such a hot topic I believe a short historical excurse is 
in order.

> Where is this policy? (url please!)
The policy originated way back when metadata were only discussed. I believe 
most general and strict formulation was in one of Daniel (Robbins)'s emails 
back then. Unfortunately many procedural "regulations" were not properly 
written down (glep's were not around either), so best coverage would be in 
the mailing list archives which we don't have, least for gmane.org ones if 
they go that far back. 

This particular point, however, is covered in devrel handbook, metadata page:
http://www.gentoo.org/proj/en/devrel/handbook/handbook.xml?part=2&chap=4#doc_chap2

Particular citation (skipping tags):
"There must at least be one herd subtag. The contents of this tag should be 
the name of be a herd as specified in the herds.xml file. It must occur at 
least once. Besides being member of a herd, a package can also be maintained 
directly."

In other words:
every package *must* belong to a herd (and may belong to multiple herds)
Additionally a package *may* have a maintainer

The rationale being (and that was touched upon in those old discussions 
multiple times) that there always is a fall-back. Thus strictly speaking 
no-herd is against the policy as this is not a valid herd (or even not a herd 
at all).

Observe also that no-herd is nowhere to be found on that page. This is not an 
omission but rather a "simptome". There was no mention of no-herd entries 
when the policy was developed and metadata format worked out. no-herd entries 
started to appear later on without much fuss. They were noticed and the issue 
brought up and according to what I remember no real decision was made on 
them. The general feeling I have is that they were "tolerated" by more strict 
policy followers with the understanding that this is a temporary measure and 
we should try to minimize their use and get rid of them whenever possible. 

Of course nothing is more permanent than a temporary solution :). So, these 
are the practical considerations I cited in the opening. One possible way to 
avoid no-herd entries is to create a more generic herd that would accommodate 
various "smaller breads" that did not warrant a herd of their own. Multiple 
developers may sign up. Individual involvement may be regulated by the <role> 
tags in herds.xml and/or <maintainer> entries in metadata.xml (observe 
and/or, this does not involve data duplication).

Again, issue at stake here is insuring that no package remains unsupported in 
the case of its maintainer departure, - a holy grail of our herding attempt 
which, sadly, is not likely to be finished any time soon.

Ok, I hope this explains situation a bit. To completely close the issue I just 
want to (again) note that main emphasis of my original message was to bring 
attention of maintainers of packages that could fall under lang-misc. 
I modified metadata of a few such packages replacing no-herd with lang-misc 
(of the latest erlang comes to mind) and removing maintainer entry listing 
myself (I am on that herd). However I left the 2nd maintainer entry intact 
and did not add these devs to lang-misc in herds.xml (which I would though 
moderately suggest ;), although I believe this should be at the discretion of 
the involved dev).

George



--
gentoo-dev@g.o mailing list