* raster@xxxxxxxxxxxxx wrote:
> actually i see zero use for pre-sorted files - it just puts complexity
> in fam that wont cover all cases - then fam has to also sit and wait
> till it's read all 1000 or whaetvere file sfirst then sort THEN dump to
> the app - no - i think it's bad.
Only the first time. After that, the cost drops significantly, and takes
the load of sorting and holding onto all of that information off of the
application, saving resources and traffic.
Since fam is designed to run (essentially) all the time, only the first
access pays. Since users generally access the same directories all of
the time, the cost is amortized well.
> it shoudl be up to the client to figure what it wants to do with the
> files it gets. it can so anything it wants then and save
> pre-processing the list which won't save processing - it just shifts
> it to fam and then limits it to what fam can do. :)
I don't see how it limits fam.. fam can maintain the list pre-sorted to
the last sorting request, or by filename (which I'm going to take as the
most common sort on the files), and simply resort as needed, with
whatever optimizations. Fam is already tracking all of that information
and caching it, it costs less to leave it in there and have it maintain
that information, sorted, system-wide, for the application. After all,
that's one of the main reasons to use fam, having one program doing the
file polling on behalf of all fam-using applications.. otherwise, you're
just inducing a ton more copying across a connection than is necessary
in most cases, for at best a small gain.
--
Source code, list archive, and docs: http://oss.sgi.com/projects/fam/
To unsubscribe: echo unsubscribe fam | mail majordomo@xxxxxxxxxxx
|