Re: client-server-IPC

[Wolfgang Jaehrling]

On Tue, Dec 03, 2002 at 02:39:06AM -0800, Frank Saar wrote:
> Hm, well I did.
> But in the meantime I succeeded in getting a working version (thanks to
> marcus) and the server is running now as a translator. But in the beginning
> I didnt want to run it as a translator. Instead I wanted the serverprocess
> to run in the background as a daemon (you know via server&) because (a) I
> was used to that way of thinking and b)) I thought that a translator has to
> follow a certain interface which is dictated by the library it is based on.
> E.g. in your hacking guide your last example was based on libtrivfs and I
> thought well thats ok as long as one wants to do the things that are
> possible with it e.g. to write to it, to read from it a.s.o. But I wanted a
> general interface I mean I wanted the client to call any kind of functions
> that do any kind of stuff which is specific to that server. But as I saw
> this was a misunderstanding.

You can just pass something else than the default `trivfs_demuxer' to
`ports_manage_operations_*' if your server implements another
interface.  If you want to start your server manually, but initially
contact it over a node in the file system, you would have to attach it
manually to that node, which - as Marcus pointed out a while ago in a
reply to Marco in a discussion related to the new console - is
perfectly possible, but means duplicating what settrans does already,
so usually there is no point in it.

Cheers,
GNU/Wolfgang