Todo for 'portal', programmers perspective - Mailing list pgsql-www

Greetings!

After having a bit of discussion here and after getting acquainted with
the source code of the "next generation" postgresql.org in pgweb/portal
I want to give some ideas of mostly architectural and political nature.

To prevent the questions like "who the f*ck is Alexey": I am a PHP
programmer, a member of the PEAR[1] community, with several opensource
packages released[2]. I also have experience programming the sites that
are  more complex than postgresql.org. Unfortunately these are
Russian-only, so no real use giving links here.

First of all, I'd like to thank Andreas Grabmüller for all the work he
did on portal module. The code *is* working and that means we only have
to tweak it a bit and not to write everything from scratch.


Now on to tweaking.

What is IMO really needed right now is removing the obstacles and
bottlenecks of website development. Consider the following examples:
1) If all the static content (i.e. the content that does not change
often) is stored in the database and is editable only through
web-interface then there is a bottleneck: the person who gives access to
web-interface. If the interface is not robust enough to let
not-quite-trusted people use it, then the bottleneck is even more
narrow. On the other hand, if the content is available in public CVS,
then every one may check it out and edit it and later submit for inclusion.

2) If some of the critical data --- DB schema, docs, ToDo lists --- is
missing from CVS then the person wishing to participate in the
development will not be able to do this without fishing for the
appropriate info somewhere.

3) If site's layout is kept in the spaghetti of PHP and HTML and even
duplicated in several files then the person wishing to tweak the design
or to contribute the completely new one will be unable to do this. The
bottleneck is again the person with the knowledge of this spaghetti.


* I18n and l10n issues

There are separate internationalization tasks:
1) Translating the framework itself and static content

For the pages that are mostly the same in different languages and for
the PHP scripts themselves gettext extension or something like this can
be used. Current solution with {LNG_whatever} placeholders is extremely
fragile, as all translations should be updated manually in several places.

Text-intensive pages with not much layout must be kept in CVS in all
availbale languages.

The benefits of this approach: people wishing to translate website will
be able to checkout the current code and work on the website without
having to ask anyone.


2) Making it possible to publish dynamic content in several languages

This is mostly done, but not actually usable without robust admin interface.

For Dave, on documentation: PostgreSQL documentation is being done in
DocBook[3] this is SGML (or XML) based format that does not have
presentation markup, unlike HTML. It is converted into other formats via
special stylesheets. This is how PostgreSQL's HTML and PDF docs are
being generated. If there is a need to translate the docs, then DocBook
source should be translated, and there are special tools for this. You
can consult PHP documentation HowTo[4], it is a good introduction for
beginners.

* Separation of HTML from the actual code.

This is quite simple: if someone wants to tweak the website's layout, he
must be able to do this without having to bother with the actual code.
Thus HTML-only templates (or with *minimal* PHP) are needed.

Very few templates will be actually needed: one for the whole website
"frame" and one for each distinct page that is available now in "www"
module, of these some will be language specific.



Now lets move to more technical issues

* Static page generation

This is working now, I suppose. I only want to suggest using
pagename.lang.html instead of lang/pagename.html to be able to use
Apache's content negotiation[5] on static website mirrors, as suggested
by Peter Eisentraut.

* Administrative interface

There is fair amount of work needed here. There should be builtin
authentication and access control system here, so that people having
access to e.g. German translation couldn't f*ck up anything else.



I think to roll out a "new" website, the following should be done:
1) Separate HTML from the code.
2) Decide upon the translation infrastructure.
3) Move all the static pages from "www" into "portal" in the form of
HTML-only templates.

That's all, the site can be published.

As soon as the people send in translations of static content we can add
them to the site. As soon as a more robust admin interface is in place,
people can be given rights to publish news, events and surveys in their
languages.

As soon as someone can come up with a professional-looking design
(sorry, Dave) the current one can be easily replaced.


[1] http://pear.php.net/
[2] http://pear.php.net/user/avb
[3] http://www.docbook.org/
[4] http://www.php.net/manual/howto/
[5] http://httpd.apache.org/docs/content-negotiation.html