The Perseus Digital Library has an extensive collection of Greek texts,
using the TEI; see http://www.perseus.tufts.edu or one of our mirrors,
http://perseus.csad.ox.ac.uk and http://perseus.mpiwg-berlin.mpg.de.
We have some experience with all the questions Dr. Solopova raises.
1. All our Greek texts are entered by a data-entry firm. We tell them
to use Beta-code; for Greek quoted within a non-Greek text, they also
tag the Greek section, like this:
Homer says, <G>mh=nin a)/eide, qea/</G>.
(We use phony tags in the data-entry stage because we pay by the
keystroke; it's easy to convert all <G> to <foreign lang=greek> when we
get the text back.)
This works very well for us; the company we deal with has even managed
Greek quoted within early modern printed books, in those difficult
handwriting-like early modern fonts. Recently, we had occasion to write
instructions for a new data-entry shop; you can see them at
http://www.perseus.tufts.edu/~amahoney/appian_job.html. The only
difference between these instructions and what our regular shop does is
that we suggested the new group could use a font whose encoding closely
matches beta-code, so they could see Greek on the screen as they keyed
2. We have been using Beta-code for almost 15 years -- back then, it
was the obvious choice and Unicode was science fiction. We find it
works nicely for ordinary literary Greek, which is most of what we deal
with. It's not good on the special symbols you want in a papyrus or an
inscription. I have high hopes for Unicode, but on my Linux desktop
system, right now, beta-code works and Unicode doesn't.
3. What we do for display is perhaps a little clunky, but it works, and
there are a couple of other classics projects now using the same scheme
(the Stoa, http://www.stoa.org, and the Bryn Mawr Classical Review,
http://ccat.sas.upenn.edu/bmcr). As we convert our SGML or XML to HTML,
we put a <G> tag where we see the "lang=greek" attribute in the
original. A filter program then runs over the output to convert
anything in <G> tags to the appropriate encoding for the display font
the user has selected. See the Perseus font help page,
http://www.perseus.tufts.edu/Help/fonthelp.html (or at one of the
mirrors) for how this looks from the outside. At Perseus, these filter
programs run as part of the XML -> HTML conversion stream; at the other
sites I mentioned, they run as a handler for HTML files in a particular
directory. There is a main driver that figures out which font to use,
based on a browser cookie, and then calls the appropriate filter
routine. These sites are either writing straight HTML or converting
SGML to static HTML files, rather than doing this conversion on the fly
as is done at Perseus.
I hope this is useful; I can provide more details, including the code
for those Greek filters, if you'd find it useful.