2001.01.12 22:52 "Re: What is the proper number of bytes in a TIFF 6.0 "word"", by Dave Abercrombie
The TIFF 6.0 spec uses the word "word" a little losely. In the description of an Image File Header, bytes 4-7 are an offset in bytes of
the first IFD, and then goes on to say that the IFD "must begin on a word boundary". But there is no mention there of the size of a "word".
Shortly thereafter, in the description of an IFD, bytes 8-11 contain the
Value Offset. "The Value is expected to begin on a word boundary; the corresponding Value Offset will thus be an even number." I suppose this
answers the question, but if a "word" were four bytes long, it would still be even.
If I'm not mistaken, a "word" was a dominant term in the days of the 16-bit architecture; and C had to make the distinction between a "short"
and a "long" (word). Since TIFF 6.0 was finalized in 1992, I suppose one could presume that words were predominantly two bytes at the time.
What is the consensus among the TIFF experts?
If you look at figure 1 (page 14 of the spec), I think it is clear that a "word" is meant
to be a 2 byte quantity, though the spec doesn't define this term.
What is your interpretation of clear. The "word" alignment corresponds to the value termed "A" in Figure 1, but there's nothing to show that "A" is aligned on a 2-byte boundary.
If, however, the fact that all parcels of information seem to be in 2-byte blocks, then you might be right. But one could only imply from the fact that all of their little boxes are 2-byte boxes, whether filled in with solid or dashed lines.