On 5/6/05, Chad Perrin <perrin(a)apotheon.com> wrote:
That sounds great, but it raises some issues that are,
for the most
part, specific to Wikipedia and not many other projects/companies/et
cetera. At least, it potentially raises such issues, depending on what
exactly you mean to do with that read-only copy. Do you mean to have it
in operation as a way to serve pages concurrently with the primary
setup, or do you mean it to be sort of a failover backup? If the
latter, I'm all for it, if the Powers That Be decide it's in the stars
for us. If the former, however, I might bring up those potential issues
to which I already alluded.
The primary use I was thinking of was as a failover option, even if
all normal servers blow up, go down, etc. But as long as such a system
is set up, however, why shouldn't it be available all the time? For
instance, if I were to design a game that involved regularly querying
WP content (and needed quick response times for it not to slow down
gameplay)... it would be much easier to do that through a central
static site than by setting up an incrementally-updating local mirror.
No need for us to rely on third parties like factindex to provide fast
static mirrors of Wikipedia content when that is what 90% of our
visitors want to see. (Or perhaps there is such a need? Depending on
said potential issues.)
Timwi: having different langs in different data centers wouldn't make
it impossible to have a cross-language watchlist; it would just have
to be done carefully, without too much between-datacenter latency...
--
+sj+