November 2008 Archives

I wasn't really sure that I'd have the need for document management in the near future when I reviewed Knowledgetree just a few weeks ago. But after a recent rapidfire of document production that flooded many inboxes, I decided it might indeed be time. Another document was on the horizon, and I saw the potential of another flood.

The competitors were Knowledgetree and Alfresco. I'd reviewed both of these before, but at the time I was willing to overlook some minor annoyances as things I could work out if I needed to put them into production. Well, it was time for production and I would soon find out that it would take 80% of the installation time to get them that last 20% of the way there.

I actually found another good blog comparing the two systems about a year ago. That reviewer gave the nod to Knowledgetree, but they had both come a way in a year and thought it was worth taking my own look.

Most of the pros and cons from that reviewer still hold true as it turns out. Alfresco is more powerful, both in the features it offers through the web interface and the ways through which you can access it (CIFS, Webdav, Sharepoint, etc.). I think for the "enterprise" Alfresco is likely to be a better choice for these reasons. But for the non-enterprise Knowledgetree holds its own.

The VMware appliance that I installed last time had limitations. It couldn't be upgraded beyond 3.4.2 (at least as far as I could find) and Knowledgetree is now on 3.5.4. I couldn't find where all of the files were, and in their effort to make things easy, not all of the instructions for generic Knowledgetree installation and administration held true. I then went to the linux installer version. But that too had complications. Most importantly I couldn't get it to play well with the headless version of Openoffice. I'm also not a big fan of linux installers as I don't know where all of the files it installs go. This was less important in this case, because I was installing on a simple JeOS virtual machine. JeOS is a stripped down version of Ubuntu and is a great starting place for making simple vmware appliances. There are some very basic starting points, from there it is easy to install ssh, and do a LAMP install via one command.

Still, I thought a source install might be the way to go. I found some good Debian instructions which worked well enough for JeOS. Still I had some problems. The first was solved when I realized I hadn't install cron in the very basic JeOS. There were some other missing packages including some php5 libraries (the CLI library and the curl library). But I still wasn't having luck with the Openoffice headless server (which is required for indexing documents). For that I eventually install some more openoffice packages including a conversion library, a java library, and the writer application. I also added localhost.localdomain to the etc/hosts file.

Finally everything seemed to work. Indexing was good and the interface is pretty basic. Webdav reading works for the most part, though I haven't tried writing. I like that it is in PHP (for the most part) which makes troubleshooting pretty easy for me.

Alfresco was a similar story. I actually installed the latest stable nightly build from the Labs edition, since they are very near a new release. The first part of the install was again very easy. There were some good instructions for installing on Ubuntu which worked on another JeOS machine. I applied some of my Knowledgetree install knowledge to getting the headless openoffice working, but Knowledgetree still complained. It wasn't until I uninstalled openoffice, tried to install openoffice 3 and then reinstalled openoffice, including the "headless" package. And finally I got it all to work. Still there were bugs, and some parts of the interface that while powerful would likely confuse some more novice users. Troubleshooting was a pain (granted I'm using an unofficial build). I still don't full understand the share/alfresco dichotomy. CIFS seems good, but given that I run the server on a virtual machine with a private IP, it was going to take a lot of work to get it seen by the outside world. Reading the Alfresco forums showed that in general modifying ports or forwarding behind firewalls was non-trivial.

So now Knowledgetree is in production, and I'll see how it goes. Check in and check out and straightforward and there are good options for "subscriptions" for email notification. The one thing that I would like that I haven't seen, which Alfresco had, was the ability to create a new document from within Knowledgetree. It would be useful to create simple text documents with notes from within, but it doesn't seem like a showstopper for now.

It has been quite a while since I posted here on the Obsessive Collaborator. This has been due primarily to the migration of my server to a new machine (alas, it was not purchased with bailout money). I thought it was time to retire my five-year old server, at least from the front line, and replace it with a new server that would be set up to run virtualized servers. This would allow me to both test many server applications in isolated environments, and allow me to regularly backup and move offsite full functional snapshots of my servers (I am perhaps more obsessed with backup than I am with collaboration).

The process of migrating was fairly slow (mostly due to the fact that I did my work in my spare time). I was able to keep the old server running side by side until the final switch which took about an hour. While there are ways of moving from physical to virtual servers, I didn't think that they were worth the effort. Instead I started from a new CentOS 5.2 image that I created, built up the services I needed and then rsynced over the directories that I needed moved. I also used webmin to migrate settings for many of the services I needed.

The base server is also a CentOS 5.2 bare bones image. I probably could have used something lighter weight, but I now CentOS and stuck with that. Mostly that OS needs to run vmware and lots of backup scripts. I tried some premade backup scripts for vmware images, but they didn't work real well with vmware 2.0. I wound up writing my own script that progressively has each virtual machine suspend, copies the whole directory elsewhere, restarts the server, tars and gzips the directory and erases the backup directory, With a fast eSATA external drive the copying is done fairly quickly.

Vmware server 2.0 itself is quite good. I waited to go live until the official release came out. It is as easy to use as the older vmware server, the major difference from my perspective being that all management is done through a web console. For the most part I like this, making managing servers and even running console sessions quite easy. It will install firefox plugins or activex controls to run the console sessions wherever you'd like. Unless that place is a Mac, which mine often is. So I need to default to a windows or linux host for running those console sessions. I considered the now free Vmware ESX instead of server, but it is very temperamental about the hardware it runs on, and the server I built wasn't up to it (it wouldn't recognize the onboard sata).

Now I'm exploring the world of vmware images, and am disappointed when an app I want to test doesn't have an available appliance. Fortunately there are many to choose from (and I also have a bare bones image that I can build up from, and will also build a new Jeos image to make it even lighter weight for applications I want to keep).

In my search for a Document Management solution, I've looked at Alfresco over several iterations. It is quite capable, but I'm not thrilled with the UI, and the open source version has an unclear relationship with the enterprise version. I have read a lot about KnowledgeTree, and I thought I'd give that a try.

I found a vmware appliance from Rpath. Rpath has quite a number of good vmware appliances. Their documentation is somewhat sparse (as has been discussed in various forums), but most appliances now have a wiki page on their site. I found this wiki page early on, and it gave some basic setup information for Knowledgetree. But it wasn't until I found the separate wiki page over at Knowledgetree that I was able to get everything configured.

Once I got that documentation configuration was better than easy, it was easy and filled with options through a web interface.

I was able to make changes to the appliance itself (like network settings) and configure the major Knowledgetree options that normally involve messing with a config file.

When all was set up, it was a pretty straightforward operation. Log in to the site and you can upload documents, create folders, move documents, provide permissions for documents and send email alerts about documents to others.

One of the things I liked about Knowledgetree was that while it supported advanced features like checkin/checkout, workflows, etc. it also seemed like it could just be a simple document repository. Making it even easier was the ability to map the directory structure to a webdav drive. Once I actually followed the instructions, it was trivial to get the webdav working from a mac and linux clients (I didn't try windows).

I may or may not used Knowledgetree. I haven't decided whether the document management is so much better than other integrated packages I use to warrant it. But I'll keep experimenting.

A few more notes on recent upgrades. Elgg is now at 1.1 and has a lot of community plugins going for it. Opengoo just went 1.0 and looks pretty slick. And I'm looking forward to the upcoming release of Liferay Social Office. On the negative side, I've pretty much given up on Peopleaggregator. Their upgrade cycle fell short and there were too many hacks that I had to apply. Elgg looks like it is filing that open source social networking space just fine.