[gopher] Re: GopherFS implementation questions

[Stefan Rieken <StefanRieken@xxxxxxxxxxxx>]

On Sat, 12 Jan 2002 19:37:23 Timm Murray wrote:

> I've been thinking about having a full-fledged Unix filesystem
> over Gopher. Something that would include uids/gids and could be
> used as a replacement for  NFS.

Hmm, this is somewhat the other way around, but is also quite cool. I
already see this email thread falling apart into little subthreads now :-)

Of course as a server, this would work just OK because you can make up your
own filesystem -> gopher policies. So an exported filesystem would be
interpretable by all kinds of gopher clients, and additionally a
special-purpose "think NFS" client would be able to interpret the data from
this "think NFS" server in a "think filesystem" way ...right?

It *would* be fairly well doable to implement this "think NFS" client.
Actually, all John's and my problems would be gone then. But it would
*only* be able to communicate with a special purpose "think NFS" server.
Otherwise, all problems are up again...

> The nice thing about using Gopher+ as a filesysem is that you can
> have full-fledged MIME types, not just relying on the filename
> extention.

"That is *so* `80's!"

My client now mostly resolves MIME stuff by a combination of an "clicked
URL history hack" (in which the filetype byte of the directory listing in
which was clicked is stored so that when the new URL will be loaded it can
be resolved) and "sniffing" the files themselves (using standard GnomeVFS
functions). Of course metadata would be better. Extentions suck.

I was wondering: according to the Gopher+ specification, one file could be
available in different mime type versions, but listed only once in the
dirlisting. Does this happen in practice?

Greets,

Stefan
-- 
"'Abnormal Psychology'. Are you a doctor, mr. Murdock?"
"Uh, no ma'am, I'm insane."