AWARE SYSTEMS
TIFF and LibTiff Mail List Archive

Thread

2004.09.08 02:15 "[Tiff] TIFF Property Sheet Shell Extension beta", by Ross A. Finlayson
[...]
2004.09.16 19:12 "[Tiff] BigTIFF Tag Value Count issue", by Joris
2004.09.16 19:23 "Re: [Tiff] BigTIFF Tag Value Count issue", by Frank Warmerdam
2004.09.17 00:09 "Re: [Tiff] BigTIFF Tag Value Count issue", by Bob Friesenhahn
2004.09.19 12:31 "Re: [Tiff] BigTIFF Tag Value Count issue", by Andrey Kiselev
2004.09.19 16:43 "Re: [Tiff] BigTIFF Tag Value Count issue", by Joris
[...]

2004.09.16 19:23 "Re: [Tiff] BigTIFF Tag Value Count issue", by Frank Warmerdam

Joris,

My suggestion is that the tag size (aka count field) should be only 4 bytes mostly because I don't want people writing extremely large tag values. Typical libraries (such as libtiff) will actually load all the tag values into memory as part of reading the data.

I agree with the various points and counter-points you raise. They match up with the sorts of arguments I have with myself with looking at design decisions like this. On balance I feel there is no compelling reason to support tag values larger than 4GB and reasons to believe that including such large tag values in a file (even a ways into the future) would amount to a DOS attack on the app that tries to read the file!

I *do* think that we should allow and support individual tiles or strips larger than 4GB mainly because some folks still like to write their files as one big strip and I would hesitate to forbid this. Well, I would *like* to forbid this but I'm not sure it would be wise. So I do think the TileByteCount and StripByteCount tags should be 8bit integers.

Best regards,

-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam@pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | Geospatial Programmer for Rent