From mboxrd@z Thu Jan  1 00:00:00 1970
Date: Thu,  9 Mar 2000 08:50:41 -0500
From: rob pike rob@plan9.bell-labs.com
Subject: [9fans] Re: 9p question
Topicbox-Message-UUID: 9ec2f956-eac8-11e9-9e20-41e7f4b1d025
Message-ID: <20000309135041.ADwV8IAtLjSXb6Qu9x7y2ueJ3JP7oknRKwV4P2vSIcQ@z>

> The real reason for the "lengthy" conversation is the Tclone/Twalk
> part.  That is part of the price that is paid to remove the '/' as a
> directory separator

No. The real reason is that after each walk the client must check whether
the point-so-far is in the mount table.  That's why it's done a path
element at a time, and why it's so slow.  Other designs have been
proposed that would allow a walk message to contain multiple
elements (requiring the server to parse and understand '/'), but
it's such a pain to change the protocol.

Seeking on a directory is forbidden because it's hard enough to
implement reading a union directory without seek. The internal
structure that must be maintained in the kernel was deemed too
hard to maintain other than by sequential access, so we made
seeking on a directory illegal.  It didn't seem worth the implementation
overhad.  I still feel that way.

>There are a lot of strange things like:
>     delim $$ define lbr ' roman "{" ' define rbr ' roman "}" '
>or
>     $CH sub c$
>which should be "CH <sub>c</sub>",  I guess.

Would you have noticed this if it didn't contain your name?

-rob