DAVFS is very slow, now.
Is there any body have good idea to improve
No Buffering before deliver the data,
just like a web cache does.
Implementation wise i dont know how to achieve this.
I think davfs could be made very small and simple
to maintain if the buffering and caching code is completely removed.
But i guess it will feel sad to remove a lot of work :(
Then davfs could also easily be ported to another OS, KDE, GNOME etc.
caching and buffering could be handled by external program (web cacher)
caching and buffering is an art by itself and require much tuning which is hard to do by the users if hidden inside davfs.
I don't think removing caching entirely is the way to go. The Windows client Xythos ships with our server caches, and it's not too slow, for example. I don't know how it is implemented, though, since we license it from another company.
Another idea would be to have a plug-in architecture for the caching component, then we can experiment with different caching plug-ins?
We could have a default - perhaps no caching - and the current caching code could be plugged in at run-time if you want. Also, that provides incentive for people who want to experiment with other caching ideas (such as using an external caching library), that they won't be forking, they'll just be creating a new plug-in.
>We could have a default - perhaps no caching - and
>the current caching code could be plugged in at
>run-time if you want. Also, that provides incentive
>for people who want to experiment with other
>caching ideas (such as using an external caching
>library), that they won't be forking, they'll just
>be creating a new plug-in.
Also I need to study about the cache itself.
What kind of cache library could we use?
Are you familiar with file system cache?
The easiest way to speed up things is to implement
as a | (pipe) what is today ; (sequential)
1. open... GET and save to local ache
2. read... read from local cache
open... GET and save to local ache | read... read from local cache
Log in to post a comment.