On Mar 28,  2:16am, pooka@iki.fi (Antti Kantee) wrote:
-- Subject: Re: CVS commit: basesrc

| On Tue Mar 27 2001 at 23:06:37 +0000, Christos Zoulas wrote:
| > In article <20010327230316.CA5782A2A@orchard.arlington.ma.us>,
| > Bill Sommerfeld <sommerfeld@orchard.arlington.ma.us> wrote:
| > >> Well, the real fun part is of course that we'd actually need to run the
| > >> _target_ file binary to create a compiled database which is usable on
| > >> the target.
| > >
| > >Eek.  The binary format should be MI (e.g., network byte order, sized
| > >types, etc.)
| > 
| > And it is.
| 
| I built the database on sparc and alpha, ran both through strings, and
| took a diff and there's stuff like this in there:
| 
| -KCAH
| +HACK
| 
| Something is wrong somewhere (maybe in me? ;)

Well, the databases are endian neutral, not big endian or little endian
for efficiency. I.e. building on the sparc will produce a big endian
database, building on the i386 will produce a little endian database.
Both the sparc and the i386 can read both databases. But if you have
a network of i386's why you should spend extra time byte-swapping?

christos