On 10/25/2014 01:50 PM, MZMcBride wrote:
> [...] that probably doesn't mean investing in Labs, exactly.
> Not if you want to have a long-term, substantive impact, in my opinion.

I'd like to address that particular recurrent canard here, if I may.

Things that reside in labs are empathically /not/ second-class citizens
by any stretch of the imagination.  Perhaps our attempts to emphasise
that "Labs is not production" were not clear enough about what me mean
by the distinction - and because of that people have gotten the wrong
impression about it.

What "not production" means is simply a matter of (a) scaling and (b)
service level.  For the latter, all it means in practice is that if
something in labs breaks not all of ops will drop what they are doing to
attend it as we would for prod.  It doesn't mean that we don't care that
it broke, nor that it is of lesser importance - just that the impact is
lower and therefore it is not reasonable to divert all resources to the

As for scaling, it will almost never be an issue until something becomes
used frequently by a large fraction of the projects' user base.  Labs
remains a perfectly reasonable permanent home for anything that expects
light or medium use - whether it's volunteer-driven or WMF-driven
(deployment-prep is an excellent example of a service that lives in Labs
which is used continually by almost all the devs and yet will never live
in "prod").

Labs isn't a second-grade production for unimportant things; it's a more
flexible, more open environment for general tooling and development.  If
anything, it's /prod/ that is more restricted (in who can use it, how
complicated it is to be allowed to deploy there, what restrictions are
place on what is there).

Any GLAM tools would almost certainly live in Labs - whether it's been
developped by the WMF, volunteers or Chapters - not because it's not
"worthy" of production but because trying to make it into production
services would make development and deployment immensely more
complicated and much less flexible.  The question shouldn't be "Do we
need to invest in Labs" but "How to we avoid the trouble of having to do
this in production".

-- Marc

Wikimedia-l mailing list, guidelines at: 
Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, 

Reply via email to