Re: File names and URLS

kevin@scic.intel.com (Kevin Altis)
Message-id: <9307161745.AA15552@rs042.scic.intel.com>
Date: Fri, 16 Jul 1993 10:43:28 -0800
To: www-talk@nxoc01.cern.ch
From: kevin@scic.intel.com (Kevin Altis)
X-Sender: kevin@rs042.scic.intel.com
Subject: Re: File names and URLS
Status: RO
>As a user, I want to know whether there are any inherent conflicts in WWW URLs
>and file names. On our Unix system, there does not appear to be a major
>difficulty, since file names can be pretty long. However, we also support IBM
>PCs, and they share some of the same fileservers. This causes us problems when
>we try to access the same files on the IBM network - because names longer than
>8 or with extensions longer than 3 get mapped into other filenames.
> (My true feelings about this are barely printable....
>      IBM get your act together....)

>It seems to me that if we use short file names, with 3 letter extensions, then
>there may be some chance that we could use single copies of files, without
>having to develop complex aliasing mechanisms --- I'm sorry about this - why
>the world has to conform to the lunacies perpetrated by IBMs continued attempts
>to sell software systems developed using obsolete notions of file names .......

There is a short term and long term solution to your problem. The short
term solution is for a PC server/browser to generate an 8 character
filename with 3 letter extension to match the longer name and then either
maintain its own internal database of filenames or generate names on the
fly. Quite a few software packages already do this, though it has its
problems; if the algorithm used to generate the filename.ext name comes up
with a duplicate name for a given directory then it truncates filename to
filenam1 and so on, which isn't very pretty and means the same file might
have a different name when moved to a different directory.

The long term solution (next year) is to wait for Windows 4.0 aka Chicago
aka Windows NT Lite which uses the same file system as Windows NT (so say
the industry rags ;>) so you get 256 character filenames  and then the DOS
FAT compatibility module of the OS generates the filename.ext for you just
in case some DOS weenies try and access your file system.

Mac people of course would like names to stay under 31 characters which is
the Mac filename limit, Mac names also can't contain colons, which is used
to separate directory names like "/" under Unix and "\" under DOS. In
general, you wouldn't want to use real long names (approx. 50-256
characters) anyway because names would then start looking like sentences or
even whole paragraphs, wrapping on the screen and prone to many typos by
less than perfect humans.

Hopefully, URNs would allow you to have many name variations for the same
item otherwise the problem you're having with DOS will be forever with us.
The same file is going to end up having a French name variation, Kanji
variation, German variation, OS variation, etc. If URNs won't solve this
problem, then we probably need to agree on some algorithms to consistently
map from the URLs to a variety of different OSs and maybe languages as
well.

ka