[sw-l] Sign Variants via SWML

[dparvaz at MAC.COM]

Sandy's suggestion should make searching for signs interesting (at
least in XPath), since some signs will have the "variant" attribute and
others won't. And since you can't predict which nodes have the
attribute you have to somehow say "in searching, make sure that
everything in the sign either has the same value for "variant"
attribute or no such attribute at all." In other words, how would one
within the DOM find the gloss for "flatfish", given the correct fill?

It seems simpler to me to list each variant as separate signs, so
something like this:

<entry>
<variant head="yes">
<sign>...</sign>
<semantics>...</semantics>
</variant>

<variant head="no">
<sign>...<sign>
<semantics>...</semantics>
</variant>
</entry>

I'm punting here, and of course, the content of the "semantics"
structure is dependent on whatever the application is (dictionary?
Machine translation?). The "head" attributes potentially tells a
lexicographer which variant(-s) should be considered as the head of an
entry in the dictionary. And searching for variants as signs becomes
much simpler. Once a single sign is found, either it can be dealt with
on its own, or other members of the entry can be found by moving up the
tree to the nearest "entry" node.

This of course assumes that sacrificing storage for more optimal
searches is acceptable.

Cheers,

Dan.