[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Proposed Text for RFC2445 Issue #1
On Tue, 3 Oct 2000, John Stracke wrote:
John, you said you didn't feel strongly about this issue :).
> First, note that this is not a completely solvable problem. If the attendee and the
> organizer use different filesystems, the organizer may use a filename which the
> attendee's system does not allow.
That's strictly true, but I don't know any filesystems that pervert
filenames so badly that they on longer contain any
information. "PROPOSA~.DOC" is still better than "X45GHHWD3.DAT"
> * A "fetch a file" URL, which includes a filename. (If it's an http: URL, the
> server can also use the Content-Disposition: header.)
Sometimes the URL includes a filename, but it is becoming more common to
distribute URLs that point into a document management system (I've used
LiveLink) where the filename is either a node number or is deeply embedded
in the URL.
Using Content-Disposition is a good way to do it, but only from the HTTP
Server's perspective. I think it just moves the responsibility for a
problem in the calendaring application to a system that is not controlled
by the calendaring spec. If you rely on Content-Disposition, sometimes you
will get it, and sometimes you won't.
> If we introduce FILE-NAME, then the first two cases become problematic: what if the
> FILE-NAME and the filename derived from the URL don't match? Different
> implementations will make different choices, and you'll be right back where you
> started.
Clients will only make different choices if the specification does not
indicate which choice to make.
If we add the FILE-NAME parameter, and specify how to resolve conflicts
between it, the Content-Disposition filename part, and the URL filename,
there should be no problems. To this end, I propose these rules:
1) Use the value of the FILE-NAME parameter
2) If no FILE-NAME, use the value from Content-Disposition
3) If no Content-Disposition, use the filename part of the URL
4) If no file-name in URL ( it is a cgi or query string ) make
something up.
Would this solve some of your concerns?
At my last company, it was very common to attach documents to Email, and
Eudora would put them in an attachments directory. I spent a fair amount
of time looking through that directory for documents that I had received
many months before. For these searches, the file names on the documents
were critical, so I think this is an important issue.
eric.