[sync] sync performance and P900

[Andrew Suffield]

On Wed, May 04, 2005 at 06:13:19PM +0200, Karsten Fourmont wrote:
> a few days ago I thought nowbody would use this code besides me and now 
> people are telling me they are syncing thousands of contacts ;-)

All it takes is one monkey who's been using their cellphone for
*everything* for like, three years.

> Well basically it's a shame that the horde server is unable to come up 
> quickly enough with the data. We'll have to look into that.

I don't think there's much can be done about it. It spends most of its
time waiting for SQL to be processed, and there's nothing obvious to
be done to speed that up. Even postgres isn't fast enough. Unless
there's some way to drastically reduce the query pressure, SQL is just
too slow for the job.

> Other than that the SyncML spec has a "BUSY" mechanism which should deal 
> with exactly this case.
> 
> Maybe look at
> http://www.syncml.org/docs/syncml_sync_protocol_v11_20020215.pdf

Dead link, but I found it anyway. I note that the spec allows clients
to effectively ignore this and just start over next time, under the
assumption that the server will be faster by then, so this isn't a
general solution. I don't know what the P900 does. It might be good
enough for practical purposes.

-- 
  .''`.  ** Debian GNU/Linux ** | Andrew Suffield
 : :' :  http://www.debian.org/ |
 `. `'                          |
   `-             -><-          |