COSMO Web-site: the bad parts

Cosmo web site: the parts

Note: this page can be dispayed correctly only with Firefox versions of 2014+

the meaning of

starting

The site started in 2000 as a "Hello World" statement; static and purely presentational. In time, its static content started to have some practical use:

  • some regularly updated content (like Italian and Swiss verification reports)
  • some historic/archival content (like all previous GM presentations),
  • some reference material (core model documentation, techReports etc)
  • lately, as a distribution place (observations, FieldExtra)

in a changing world

Meanwhile, the web was changing; it became more stable after the browser wars, it drifted for some time wondering about syntactic purity and currently focuses on being an application-centric platform (and still evolving at that). Even its hardware changed; from big iron to distributed virtual machines coming and going by load and locality.

trying to keep up,

The COSMO webSite slowly drifted along. Some applications appeared in it:

  • Some tables comparing the parametrization of the different centres; a 1st version of a manual of namelists and their variables
  • tracking of tasks and their FTEs, developments and releases.
  • pages that could be edited and ways to upload files came to encourage user directly generated content
  • We tried if external apps (forum and bug-report management) were of any use
  • A complementary mail server appeared for centrally maintained maillists

successfully or not

The "BAD parts" is a presentation about those pages or server-side applications within the COSMO web site, that during these years

  • were started but didn't keep up or were entirely abandoned,
  • are still there but could have fared better (in quality or content wealth) and
  • are missing altogether but could have been there

ness around

Some problems come not directly from the site but from what supports it; the system it runs on and the network connecting it to the world. Here are some:

the system

Both the web and mail server run on the same system, a 3.6GHz single-core double-thread Xeon with 2GiB memory. This system was bought in 2005 as part of 6 support PCs for the main computer and remains as the one of the two still in service. It is still performing well and is (more than) adequate for the its light load, but it has some disadvantages:

  • It is subject to non-recoverable failure. Its siblings had such deaths: a part failed and there was no replacement to be found (being old or too brand-customized)
  • It's CPU does not support native virtualization (and full emulation is too slow), so you miss the trivial replication/backup/rollback of a virtual machine

and its network

The system is placed within the HNMS firewall, so all its limitations apply. The system cannot initiate a client connection for any protocol other than SMTP (mail distribution) and as a server is blocked for most protocols (other than (restricted) ssh and http)

This causes problems:

  • Replication with its Swiss mirror is tricky. Some of it can be only one-sided (the Swiss mirror can pull updates according to its own schedule; the Athens one cannot push changes) or require tricks (like ssh tunnels)
  • Updating is trickier than it should be; From within HNMS it takes three jumps to place a single file; four from outside HNMS
  • All this protocol blocking sometimes makes things less secure. Some pages should be served under https: instead of http: and access should be controlled with annual certificates instead of a static password
  • Many useful applications have their own communication protocols or use ssh (mySql replication, git) so they are unusable
  • There are simple annoyances (like spam mail) that could be avoided

and the within

the static content,

Even sources of presentational/archival content have dried up:

  • Many regular events and their associated files won't find their way to the COSMO site (e.g. User Seminar presentations are up to 2009; Most PP/WG meetings page is blank or too old).
  • The events page does not contain all references (e.g. workshops are missing).
  • There are no more country-specific results or presentations (like test cases or testing results)

the few applications

Some generic applications were not used (at all) and finally were removed (like the user forum). Others are going this way (like the bugzilla subdir).

Some more COSMO-specific applications haven't evolved beyond their initial demo-level; notably the per centre namelist comparison and the namelists web-based documentation

for others, even more specific, there are all kinds of usage patterns:

  • PP/WG task management pages are not used at all by some or some find it too complicated
  • The developments page is sometimes overused (e.g. they contain links to downloadable documents that seem more fit for a PP page)

and the small things

Some minor-importance defects are creeping in. They are not critical (and won't be for some time (or ever)), but they are a concern. Here are some:

  • Users upload supplementary files in an a random way
  • The online editor produces poor-quality html
  • a lot of user text ends up in a database, so it is not easy fixing small syntactic and stylistic slips (like absolute instead of relative paths)
  • Only one person is nearly 100% aware of the site's WHATs and HOWs
  • Some (or all) site-related software needs replacing, but the downtime would be long

is it really that ?

Not unequivocally, it depends on the site's purpose.

keep it simple,

If its intent is mainly to be a presence, let the word it is there and just present some links to external contact points, then it is not bad at all (or even good at it)

or more complex?

If it is to evolve into a central application server for customized, COSMO-centric and COSMO-specific application, it falls short as it is today.

With "COSMO-centric" I imply two things:

not much,

First, that we cannot antagonize dedicated general-purpose sites: probably we won't ever be a video-conference site; we won't fix a final meeting date in it's pages; maybe we won't even need a generic forum in it. Some of these are not worth the trouble, even for when we can make them, because a dedicated site will perform better; and COSMO is not a consortium about web-development.

just enough!

Second, that there are applications that are so specific that you won't find them elsewhere. These can be useful, be centralized, and have a/the COSMO webSite as their user-friendly interface. Here are some ideas:

  • Could you generate a correct set of namelists for your domain by answering some user-friendly questions in a/the COSMO site?
  • Could you generate the external fields for your domain using a central repository with a web interface?
  • Could the documentation be remotely edited in a/the COSMO site and produce always up-to-date user PDFs (instead of the usual other way around?)
  • Could code distribution and development be managed by git in a single system with an optional web interface complementing its cmd line tools?
  • Could VERSUS be centrally installed for all members?