[Top] [All Lists]

Re: let's get rid of image/gif

1995-01-10 18:57:39
Also, application/postscript is similarly encumbered. Are we going to remove
that as well?

You don't have to use LZW to construct an application/postscript file.
You do have to use LZW to construct a GIF file.

This has not been shown to be true. In fact there is every reason to believe it
isn't true.

LZW is a particular methodology for picking common substrings out of an input
stream. Once selected these strings are replaced with references to their
previous occurences. On reception the process is undone using another specific
method for building a dictinary of substrings.

The method of picking substrings is known to be patented. The method for
decompressing the result may or may not be patented. (At least two legal
opinions have been obtained that it is not patented, despite claims by Unisys
to the contrary.)

It is quite possible that an alternative unpatented method exists for picking
strings that generates the same format. And it is almost certainly possible to
perform the decompression using a different approach.

It remains to be seen whether this approach will be taken in dealing with this
GIF problem. Another, similar approach is to devise a new GIF format that uses,
say, LZ77 technology that is known not to be patented. (GIF includes a version
field that makes this possible.) And still another approach is to select a new
technology (JPEG or whatever) to replace GIF. Or it may be that the licensing
terms are reasonable enough that all of this is irelevant.

This is why I recommend a wait and see approach. This fight is far from over,
and the outcome is far from certain. We will all look very stupid if we rip a
very useful type out of the MIME specification only to find that there was no
reason for doing so.

Unisys is said to believe that the LZW patent covers decompression as
well as compression.  If that were true, there might be a reason for
MIME to recommend against use of LZW in application/postscript files.

I'm sorry, but this goes too far and has no business being in a protocol
specification. We already have specifications that warn against using patented
technologies. That's enough.

The only PostScript issues that are relevant in MIME are security issues, and
to be honest I think a separate "PostScript security issues" document would be
a good idea.

But the whole notion of true/false is pretty hazy when it comes to
patent claims...

The truth or falsehood isn't hazy, its irrelevant. The real question is what
will a court say about this when asked. And since getting an answer costs big
$$$, we try to make do with legal opinions from more-or-less knowledgeable
experts (which may also cost $, but not as many). This is where the haziness
comes from.


<Prev in Thread] Current Thread [Next in Thread>