Re: [HACKERS] New INSTALL text file

Bruce Momjian wrote:
>
> I have taken Peter's new install.sgml and generated a new INSTALL text
> file to match it. I used sgmltools to convert to HTML and netscape to
> dump the html as text.
>
> File attached.
>

There are only two things I would want to see different. The
first is the example of running configure. Even though it is
beyond silly to think that people will interpret step 4
literally, I guarantee you some will, and will try to enter:

"./configure [ options ]"

You see it occasionally when users complain that they can't
create 'C' functions per the documentation:

CREATE FUNCTION add_one(int4) RETURNS int4
AS 'PGROOT/tutorial/funcs.so' LANGUAGE 'c';

People actually enter "PGROOT". On the vpnd list just yesterday
someone was complaining that the IP address of 324.240.123.122
(or something like that) was illegal and they couldn't get their
vpnd software working - despite the fact the documentation
explicitly notes that the IP address is an example and is
intentionally incorrect to prevent people for messing with
others' networks. People "key what they see" I'm afraid. So an
example configure command would be nice.

The only other thing is if somewhere there is a mention of the -o
-F options for the backend, suggesting its possible use. Since
fsync() is on by default, many people who don't dig into the docs
and are just trying PostgreSQL to see if its a plausible solution
may dismiss it out-of-hand for performance reasons. Even though I
know robustness is the #1 criteria for a RDBMS, I personally
believe fsync() should be *off* by default, but I know I'm in the
minority.

Just some thoughts,

Mike Mascari