|
|
Home | Docs | Issue Tracker | FAQ | Download |
|
|
| Date: | 2008/11/5 |
|---|---|
| Author: | Howard Butler |
| Contact: | hobu.inc at gmail.com |
| Last Edited: | $Date$ |
| Status: | Proposed |
| Id: | $Id$ |
Developers and users have expressed dissatisfaction with the current MapServer site, the site rather poorly serves the needs of the project in many instances, and its maintenance and upkeep is limited to a single system administrator (Howard Butler). This RFC will aspire to replace the current MapServer website with a hybrid setup that is similar to the infrastructure that the OpenLayers currently maintains.
The current MapServer website could be considered the 2.0 version of the MapServer project’s web presence. The 1.0 version of the website was a completely static. The current version of the website looked to allow through-the-web editing to lessen the burden for documenters. Over three years into the effort, it is pretty clear that the website has not had the desired effect with respect to documentation, and it is getting in the way of the project doing other business.
Our uptime has been ok, but one effect of the current setup is that no one except Howard Butler takes responsibility for our web infrastructure. Part of the reason for this is there are no other Plone admins that have volunteered time in the MapServer community in the three plus years the site has existed, and part of the reason is that Howard bootstrapped the 2.0 version of the site and it was easy to leave it in his hands. Howard doesn’t have the time to be able to keep things up at anything over a subsistence level, and administration of MapServer’s web presence must be distributed if we are to achieve any progress.
A survey that hoped to determine if the community had any feelings about how the site is currently serving the community was rather inconclusive. While generally positive about the site, the self-selected sample size of eighteen dwarfs the nearly three thousand mailing list subscribers, and I am uncertain what was expressed captures the general sentiment.
Here are some goals that the 3.0 version of the MapServer website should achieve:
Some people have complained that it is difficult to find documents on the website unless you know the exact place in the hierarchy. Because our mind-reading webpage finding software isn’t quite up to snuff, the new website should make it easy enough for documenters to organize and reorganize information in logical and interlinked ways. It seems the strictly-enforced hierarchy causes more problems than it solves in this regard.
The current website is quite slow. Slow to edit, slow to view, and slow to change. There’s lots of pointing and clicking involved to do the simplest tasks. So much so, that folks will only update the website when they absolutely have to. Developers have subversion access by definition of being developers. They should be able to edit the website through text files in subversion and have the website be updated automatically.
The website fails documenters in a number of ways, but the most important failure is the inability to tie documents to specific MapServer versions. A new iteration of the MapServer website must allow this to happen. Luckily, we already have tools to version documents (our source code repository), so we should just leverage that to accomplish this goal.
From time to time, users do contribute significant documentation describing how to accomplish a particular task with MapServer. Our new infrastructure must still allow this to happen without too much friction. MapServer’s Trac instance already provides this capability (along with single-signon), and we can take advantage of it to accomplish this goal.
The OpenLayers Gallery works better than the current MapServer gallery, and it works much easier from a management/administrative standpoint. A benefit of using OpenLayers’ gallery software is both projects can enjoy the benefits of improving it, which is not possible with the current MapServer gallery.
Just recently (Sept 15th – Sept 16th, 2008), the server that houses the site was having power supply unit issues (they have been resolved), but it is a fact that the site is running on a very old Solaris machine that could be decommissioned at any point without much of a head’s up. MapServer no longer brings grant monies to the UMN, and while they have been gracious to continue hosting us, we need to move somewhere where we have more control over our destiny. Reasons like this are exactly why OSGeo exists, there are resources there for us to use, and we should move the website there at the same time.
We are going to unabashedly rip off OpenLayers’ web infrastructure. This includes the gallery, static website, and Trac integration. OpenLayers’ web infrastructure meets a lot of the goals above, it stays out of the way of the developers and does a good job of serving the users’ documentation needs. The mechanics of how this transition will take place are described below: