Wednesday was very productive, though mostly for corridor conversations.
I found out first thing that I had a lightning talk accepted, so had half a day to put together slides about "Perl is too slow" (which is a controversial title, but I'm really just talking about PPerl).
I sat in the lobby doing slides and trying to get some benchmarks written for my lightning talk - this was made hard by the network (usual story - there are just too many people here on Wireless now and it's not coping - though it's fine if you find a room with less people on the AP).
While sat there Vivek Khera introduced me to D. Richard Hipp (author of SQLite) and we talked a little about the performance problems I am seeing in 3.0.3 on Mac OS X (though we didn't get very far with figuring out why). Apparently Vivek and Richard used to be old room mates in College - small world!
Then I met up with Paul Graham, who was happy to see a friendly face (we know each other from the MIT Spam Conference which he arranges) and I thanked him for his talk.
Then I met Theo and George Schlossnagle, who I had been hoping to catch up with a bit because they're just really smart and are doing a lot of the things we're doing, but better. So we chatted about a few things such as global load balancing, hardware, eBay, mail servers, poll, epoll, Ironport, embedding perl, threads, etc.
Then Brad Fitzpatrick joined us - Brad sat in on my Spam talk, and I had wanted to catch up with him to chat about some of the things he's doing at Livejournal in terms of scalability. He showed me some code he's using to do kernel syscalls from perl to improve networking performance. Very cool and directly applicable to our spamtrap. I want to chat more with Brad about LJ's MySQL implementation too, so I'll try and catch up with him today after my talk.
Heather joined me for lunch and we ate the Microsoft sponsored sandwiches (which you're not allowed to unwrap) with Rael and Asha Dornfest who we hadn't seen for a few years.
In the afternoon I sat in on Tim's DBI "Any Questions" where I had to suggest to Tim to add some bugs to DBI because he's made it too good. There were more questions about specific DBD drivers, which I think points to the fact that the DBI core is solid.
Then I practiced my lightning talk on Heather back in the hotel room, and returned to the conference to deliver it.
The lightning talks went without too many hitches (some video problems at the start) and people seemed to enjoy my talk, which I tried to give in Damian's "Punchy" style, but probably failed horribly due to nervousness (it's much harder to give a 5m talk than it is to give a 3 hour tutorial!).
After that I continued some corridor conversations, primarily with Theo - talking about mail servers as usual. I really wish I could persuade our management to bring Theo over for a day or two's training and consultation - he really knows his stuff when it comes to scalability. He also has a book coming out about it, which I figure I'll get ML to buy a few copies of.
In the evening I had dinner in the hotel room (we cooked up Tacos using our enormous cooker - I don't think it had ever been used before) with Heather, and then went off to geek out a bit.
The evening ended on a bit of a damp note - we started talking about database replication. There should be an unwritten rule at conferences: Do not talk about databases in the evening.
Thursday (today) I talk about Qpsmtpd, and then I'm all done. I've given the talk before, and though it's not a "fun" presentation, it's a nice tool and so I think it'll be interesting for some people regardless.
Let me know the details and I'll sort it
