Would it break the elf rules if all the symbol names in the
dynamic symbol table (string space) were aligned on 4 byte
boundaries?

If the symbol length were known (ie saved when the hash
is calculated) this would allow a word sized memcmp()
to be used instead of the less efficient strcmp().
(the knowledge of padding bytes after the string
would be even more useful....)

	David

-- 
David Laight: david@l8s.co.uk