[4939] in SIPB bug reports
Re: Bug in Mosaic
daemon@ATHENA.MIT.EDU (yandros@MIT.EDU)
Wed Oct 19 17:27:57 1994
From: yandros@MIT.EDU
Date: Wed, 19 Oct 1994 17:27:16 -0400
To: warlord@MIT.EDU
Cc: bug-sipb@MIT.EDU
In-Reply-To: <9410191730.AA06054@toxicwaste.media.mit.edu> (message from Derek Atkins on Wed, 19 Oct 1994 13:30:16 EDT)
To paraphrase bert:
I disagree: If I give a Web client the URL:
file:/mit/warlord/www/home_page.html
That's not a URL. In fact, if you give a client that, it is free to
remove your home directory. Remeber ``breaking requirements clauses''
from 6.170? :-)
I would _expect_ the client to see "file:" and realize it should
just read the file, and then "/mit/warlord/www/home_page.html" and
open that file and read it.
you might expect that, but you're not a web browser, and that's not
the way the web works. As I said, that's certainly a reasonable
implementation choice, as long as the other operational requirements
are met (cacheing, proxies, translation/mapping, etc.). In the World
Wide Web, ``a client or dedicated sub-client can never assume that
it's actually getting the `real' copy of the document'' (this text
might look familiar. :-)
If the client knows that it can't parse the file itself, in the case
of a MIME-type that the viewer knows requires an external viewer,
the web client should pass this filename, which is _KNOWS_ is a real
file (it got a file: URL type!), directly to the MIME-type viewer.
The client doesn't get files from a server, it gets documents that map
to URL's. Even with the `file' scheme.
If I say "file", I do not mean "talk to this server", I mean
open(filename, "r");
That hardly surprises me, since I know you for the UNIX weenie that
you are. :-) However, you are not a web browser, and this is not the
way the web works at that level. the strategy you mention is
certainly a reasonable implementation strategy for UNIX web browsers,
it is not valid to assume that all browsers function this way.
chad
