Re: Re: [ANNOUNCE] Request for speakers at O'Reilly conference

Bruce Momjian wrote:

> Interesting.  I think GIST itself may be too specific for a talk.
> However, type extensibility would be very interesting.

As it turns out, there are just two am's one will most likely build
new types upon: btree and GiST. A useful tutorial should include
examples of both, explain the interface, and probably offer the
audience help in solving their own tasks. But if I were doing that
(chances are I will -- I am going to submit a proposal tomorrow), I
would start with extensibility in general -- because, as I have
recently discovered, it is either not at all obvious to people that
one can come and add their own code to the server, or the difficulty
of doing so is perceived to be too high.

Two engineers next door stopped by the other day and asked me to show
how I do it. I had one of them take a glance at the manual and build a
working library with a string concatenator from the example. That took
them 15 minutes or so, and their final reaction was, "This son of a
dog knew about it for years and didn't tell us!". Which, of course,
wasn't true: I kept telling everyone. Then the accusation was, "You
didn't tell enough -- you failed to explain how simple it was!" The
disbelief was so great that they took the code, went home, ran it
there and were even more surprised because it still worked. It must
have been yet another RTFM case, but we as a group clearly ought to be
doing more tutorials.

> The issue is that a tutorial on it will provide free air travel, while a
> presentation does not.

That probably indicates the futility of speech as a communication
medium.

--Gene