MishMash

Plone 2 beta 2, GRUF, LDAPUF, and CMFLDAP

Yesterday was one of those days where you get home, later than you might have wished, and grunt at your housemate when he asks how the day was. I'd tried to move forward with half a dozen things, and failed with all of them.

Today has been a little better.

• I have the latest versions of Zope, CMF, and Plone up and running (zopectl was running python, despite being asked to run python2.1; crude fix: make them synonyms).
• I have LDAP working, and have worked out most of what I need to know about LDAP and LDAP schemas to use LDAP to store user info for Plone.
• I have some code working which extracts user details from the current, old old old, Plone set up and writes it out in LDIF form ready to be imported into LDAP.

None of this is exactly finished, but I feel as if some of the huge blocks I've been pushing against on and off for a while have started to slide. I need to upgrade the FloodRiskNet site urgently. It's been creaking for ages, and it doesn't do any of the community-encouraging things that I had planned for it very well. If I got it going with the latest version of Plone, and cleaned out some of my early hacks (I might even ditch some of the features I was trying to achieve with them; simplicity is king) I could probably start adding the odd extension that I've had in mind for yonks. Like a proper "publication" type.

I really want to move away from storing member details inside the FloodRiskNet site for a number of reasons. One is that doing this I think will allow me to migrate to the new version less painfully, and has the potential to serve that purpose in the future, too. Another is that I can pull all of the member details out for use in tools outside of Zope; mailing lists, for example, using proper mailing list software (at the moment I'm stuck, unable to mail the membership because my bodged MailBoxer doesn't want to play, and I can't work out why.

Another is that, for all that I really like Zope, and some of what I like about it is a direct result of the ZODB object database it uses, the fact that all of my data is stored in there really unnerves me. Irrational, perhaps, but it does.

I still haven't got the user thing sorted out, though. Plone 2 uses a GroupUserFolder rather than a straight user source. This means you can do funky things with assigning users to arbitrary groups, all through the Plone interface. I really want this, but I want all of my user data stored in LDAP. Now, apparently this is possible, but when I try it, I start getting errors. CMFLDAP doesn't know about GRUF.

I can't quite work out how I need to set up my LDAPUserFolder which is acting as a group store either.

Hmmm. There's only so much you can achieve in a day, I suppose.

Plone 2, incidentally, rocks. There are a load of nice features that I've come across already in my experimenting. I'm sure there are plenty more waiting to be found when I have the kinks ironed out.

Necessity is the mother of action

So total crash on the server running the FloodRiskNet web site has finally driven me to move it over to the dedicated server I set up for it a while ago. It's been running since the outset on a machine which is also used for heavy computation and I've put off moving it for a long time because the site is running on a rather volatile collection of pre-release Zope packages.

Near miss

I have a web application running on Zope on Debian GNU/Linux. Debian has excellent package management, with the facility to give a simple

apt-get update

on the command line to update all of the software on the system (this being Free software, the Debian project can package lots of application level software along with the core "operating system", which makes managing the software you have installed much easier and cleaner than it is in Windows).

So far, so good. Debian also has three versions available at any one time: stable, testing, and unstable. Stable is the officially released version, but Debian's release cycle is slow, so a lot of people run testing, which is supposed to be a mostly working install (unstable can be hideously broken). Which it almost invariably is.

Unfortunately, these two features -- easy as pie installed software management and the ability to choose to use the testing distribution -- means the command above will upgrade everything which has changed in the testing distribution when you give it. Which, when you are running production servers, is Wrong with a capital Wuh.

Fortunately for me, on this occasion the software in question was Zope, and it popped up a configuration screen before going ahead with the installation. Not every package does. If it hadn't, I'd have been screwed.

Moral of the story? Wake up, you daft bugger, and think about what you're doing on a production machine.