[Commons-l] Does Commons need a separate server for external tools?

[Gregory Maxwell]

On 8/19/07, Brianna Laugher <brianna.laugher at gmail.com> wrote:
> Re: the discussion about an API and external tools,
> We have a bit of a problem in that we are using both the toolserver
> and the regular servers in ways they were not intended for. They are
> both really intended for Wikimedia-internal use. If we use them for
> external tools, Wikimedia does not directly benefit. If we were to
> have any great success, it would likely backfire and result in the
> thing being shut down or restricted.
>
> So, do you think we should (sometime in the future, not tomorrow :))
> consider approaching a chapter or outside org to donate hardware
> specifically for this purpose?
> Call it the Commons thumbcache server.


I'm not sure I see a clear need at this point.

Commons traffic levels just aren't significant enough to be an issue
and maintaining an additional database replica is a substantial
labor... some of the commons tools require access to copies of all the
sites as well.

I think that for specific high traffic tools with well defined needs,
like search, there might be a stronger case. Especially if the tools
wouldn't require running another full database slave.

For general grab-bag tools thats pretty much what toolserver is for
and it provides a lot of infrastructure to make those efforts easier.

We shouldn't let system capacity be a limiting factor in our actions,
however, if we do build something that needs more resources than
toolserver can handle, I will be standing beside you asking for more
resources. I don't see any reason that they shouldn't be provided if
needed.