Re: YML - Why Markup Language...

David Long (dxlong@aol.net)
Tue, 24 Mar 1998 17:57:23 -0500


>Optimising for reading and delivery makes sense; IFF has that much
>right.

Yes, the haiku and TP:C&T reference were both supposed to be supporting
the natural asymmetry viewpoint, but with the caveat that sufficiently
short and wide pipes lead one back to symmetry for simplicity.

After further thought, though, I wonder if Adam and Rohit may have been
on to something: one can build both retrieve-friendly and update-friendly
structures out of raw bits without changing the semantics at the bit
level; are there semantically higher-level structures that work well
for both scenarios, even if particular representations are biased one
way or another?

An example would be a mechanism in PNG; it defines a class of chunks[1]
which are dependent upon the image data: if present, they may save
reader computation, and writers can either recompute or elide.

-Dave

[1] http://www.w3.org/TR/PNG-Structure.html#Chunk-naming-conventions