Hiya all,
A long time ago, I examined the MySQL client-server protocol, to understand why it is performing so well. This resulted in the idea (back then) to create a MySQL server implementation in merovingian, that proxied commands to mserver/mapi calls. This to possibly gain from performance benefits, but obviously more to open up the world of MySQL clients that exist to MonetDB. A minimalistic start was added to the code [1], but nothing more than that ever emerged.
Years later, I ended up at a similar quest, one where I had to implement the MySQL server protocol and perform the rituals to make a MySQL client believe it is dealing with a proper server [2]. Objective is different here (some will recognise merovingian's multiplex-funnel feature), but the codepath is largely re-usable for the purpose of the above.
Being rusty with the MonetDB code-base, not having followed any of its developments lately, whilst on this island in the sun, I'm wondering if it makes sense to implement the first at all, or whether it is better to simply remove the if-deffed code, for it is incomplete.
Opinions?
Best, Fabian
[1] http://dev.monetdb.org/hg/MonetDB/file/fb8e50da2917/tools/merovingian/daemon... [2] https://github.com/grobian/mysqlpq