On 08 Oct 1999, Kai Gro�johann <[EMAIL PROTECTED]>
wrote:
> Charles Karney <[EMAIL PROTECTED]> writes:
>
>> The right solution would be to move the rcp hooks up the Emacs
>> calling chain. This would allow the safe save to be implemented as
>>
>> copy REMOTE:FILE to REMOTE:#TMP#0 rsync BUFFER REMOTE:#TMP#0 #
>> potential rsync efficiencies here mv REMOTE:#TMP#0 REMOTE:FILE
>
> That would be a nifty idea. I'm wondering if anybody else is doing
> stuff like this, so you all could lobby Kyle to do it...
>
> I admit that I'm not sure at all about doing the cache thing. It's
> not clear to me when the cache should be invalidated.
I must agree here. EFS, under XEmacs, is quite fast for remote
operations through it's use of a cache for file properties, directory
contents and the like.
It also has the single most irritating inability to notice when new
files are created in a directory. RCP, while somewhat slower, works
right every time.
This is especially notable for me as much of what I do with these tools
is remote editing and compilation. Changes on the remote machine that
emacs does not start make the cache difficult to maintain.
OTOH, caching some file properties would substantially reduce the load.
It's just that, as Kai says, invalidation of the cache is a difficult
one to time correctly.
Daniel
--
Forgiveness creates a shift in perception that permits us to see our mistake
as an opportunity to learn rather than as proof of how "bad" we are.
-- Joan Borysenko
