XAPI is a very high performance solution and actually uses less resources than the normal API.  It is also quite stable.

It's current speed is a consequence of the the fact that there is only one server taking all of the load.  It is regularly overloaded.

A dedicated service for Wikipedia use would really help make it faster for everyone.

Since it is new to the Wikipedia environment it's fine to classify it as experimental for the time being - but it provides useful functionality that the normal API does not have and is necessary for applications such as Query-to-Map. 


On Mon, Oct 5, 2009 at 10:46 AM, Marcin Cieslak <saper@saper.info> wrote:
> Hello, we had this discussion in the past.
> Could we get a XAPI-instance[1] on one of our servers?
> I would like to use it for my Query-to-map project[2]. For other people
> this service is also usefull.
> In the moment  the XAPI is so slow (2 kbit/s)  that my script is
> out-of-service.

I understand that there are numerous issues with XAPI software
(being that slow, using lots of resources, etc.) therefore it should
be classified as "experimental". We should definitely watch for the progress.

Ptolemy is still importing OSM data right now. We need to find out
how to tune the database properly and what our space/speed requirements are.

Do you think you could use normal OSM API (instead of XAPI) for your projects?

> Who could decide this? Saper?

It's for us to decide, the only real problem is to assess how production-ready
XAPI is. We need to remember that Ptolemy is going to be the WMF production box.

> Other question, could cassini get more hard drive space? How usefull
> would it be?
> The concept of the the three servers usage is now other than on the
> beginning.

I don't know - what do you plan to use the extra space for?

             << Marcin Cieslak // saper@saper.info >>

Maps-l mailing list