Re: A femtolisp based emacs clone

[Michael Albinus <michael.albinus@gmx.de>]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> Indeed.  I think you could get the best of both worlds by using an
> approach similar to Tramp but running a little tramp-daemon on the other
> end, such that we can use a more efficient communication protocol
> (e.g. when saving a file we could locally diff it against the last
> version (of which the remote daemon would also keep a copy), and then
> only send the diff).
>
> This said, even using such a daemon it might be difficult to get good
> performance: part of the problem is the number of round-trips.
> E.g. when saving a file we have to check if the file was modified in the
> mean time and whether saving into a new inode would change the owner
> (etc...), which each require a round-trip.  To get rid of these
> round-trips, we'd have to shortcut this code and delegate the
> higher-level "save file" operation to the remote server, which then has
> to perform those tasks but still obeying the locally set customizations
> about how to do each one of those tasks.

A while ago, Tramp has tried this approach, using the FISH protocol
(<https://en.wikipedia.org/wiki/Files_transferred_over_shell_protocol>).
Experience was bad wrt performance, so Tramp removed this protocol meanwhile.

But I'm open to restart such an implementation, if there would be a
better protocol. Preferred an existing one, but we could also design an
own (and write then the server part as well).

>         Stefan

Best regards, Michael.