Re: Re: cvs: php4 / TODO-4.1.txt

On Wednesday, April 18, 2001, at 03:53 PM, Stig Sæther Bakken wrote:

[Ron Chmara <ron@opus1.com>]

"Frank M. Kromann" wrote:

Is there a list of modules that stays ?

A proposed list where it "cuts close to the bone": (since nobody seems to want to be the 'bad guy', I'll start...) M= Main P= Pear (To show/expose any personal bias, items marked with a "*" are ones that I use on most servers.)

Major Snip_>

pcre P

PCRE should be in the main distribution, a lot of PEAR code relies on it.

Snip Again _>

xml P

*BLAM* That's the sound of someone shooting himself in the foot. The PEAR installer needs the XML extension. :-)

LOL! Well, so we've established that for PEAR-based-extensions to work, we _must_ have some of the /ext's included in the main distribution, as well as some form of dependancy checking and resolution. Looks like we're morping into a PEAR design discussion, in some ways. :-)

Which would mean that the main distro would have: apache ftp

Why ftp, when for example curl is out?

My mental approach was to ask myself what the basic, internet-related, extensions were, the things likely to be used by a beginner before they grasped more advanced concepts, so they wouldn't have to learn PEAR interactions to get started with the bare basics.. That's why I picked fairly well-known names, packages, and databases. While cURL can do a _lot_ of things, a newbie trying to ftp, get ldap entires, etc. might be more likely to look for functions directly associated with their end goal. That was my reasoning. Not neccesarrily the best reasoning, tho. :-) It assumes a learning curve for PEAR, etc. Of course, if we have sockets, we don't actually "need" ftp, imap, ldap... well, just about anything that writes or reads :-) .... but that makes the learning curve a bit steeper. Clifflike, even.

mssql mysql pgsql

I'm willing to bet my best cap that oci8 is much more used than mssql.

The reason that I wound up with these three is that I was looking for the "top two". To compensate for our Win32 users, I added mssql, because IIRC the pg port was pretty much a cygwin-kind-of-kludgy-thing... (Of course, the WIn32 users do add a whole new dimension to this....) Hmmm.... it sure would be nice to have some actual metrics on extension use, regardless of what gets put into PEAR. Do we have any logs on www.php.net/manual page views? Could we use that in some form to determine extension use (by how frequently an extensions' pages get looked up)? It's not survey based, and it's passive enough that timeframes that we were collecting data on could be hidden from poll-slamming. Even if we didn't want to log web hits (in general) for performance reasons, (I have no idea if we do) we could at least get a set of numbers unbiased by personal use/coding styles/etc. by turning it on for a minute or two, once an hour, for a few days.

pspell

Why?

I guess I use it often enough that I associate it with being as valuable as any other form of standard string-validation. It's the hidden biases (and I guess the above is one of mine) that creep into this.... I'd be much happier if we had some metrics on use. Maybe we could also gather information on number of errata notes for each, opened/closed bugs for each, and pageviews for each. Those numbers should at least point us in a few directions, where it gives us an idea of what people are actually working with (as compared to just installing, or a web-poll.) Something that wasn't too geographically based would be nice, as well, because application usage (which interacts with the extensions used) varies from country to country, timezone to timezone. -Ronabop