I think I'm the only one except Alex (at least publicly) who actually
took time to write a useful/largish web application (vizreader.com). I
wanted to see for myself what that would entail.
So vizreader is an RSS reader which shouldn't be too difficult, read
som XML, store it in a database and then present it and let the user
read and categorize it.
Some issues I had:
1.) Avoiding ugly urls with a lot of parameters. Something like
/read/blog/10/article/5 is preferable to !read?blog=10&article=5 My
solution was working around by doing a lot of AJAX since that's easier
than writing a url rewrite lib, so...
2.) I needed JSON serialization to and from PL, wrote it myself but
it's not a releasable library, too little time and energy to make
something rock solid (at heart due to lack of capability, a master
could maybe do it in a few hours, not me).
3.) Pretty soon it dawned upon me that perhaps as much as 25% of all
RSS feeds out there will break the XML lib in PL so a custom more
forgiving solution was needed using match (still a work in progress).
Languages like Python has probably already dealt with this by already
having XML parsers that can handle quite a lot of abuse.
4.) Since Twitter is the "micro blogging" platform and also outputs
RSS I wanted to be able to read it too. However in order to login my
users I needed to observe all the rituals as prescribed by something
called Oauth and Twitter has its own version called TwOauth which was
implemented in PHP, Ruby and Python already so I opted for the PHP
version since I was already calling PHP for some small things from
PicoLisp (I didn't want vizreader to be dependent on yet another
language apart from PL and PHP). Unfortunately the PHP TwOauth needed
to be massaged into a CLI version (since I'm calling the CLI version
of PHP from PL) since there were a few CGI only things in there.
5.) I ended up downloading a c executable to do porter stemming and
calling it through PL (languages like Python already has things like
this written of course, you just google and you find).
There are probably more but that is what I can think of at the moment.
These kind of problems have nothing to do with PL in itself and
everything to do with the age old adoption problem, it's like a dating
site, the more people who use it the more utility it has and it's a
little bit like that with languages too but in a derivative way.
In reply to other things that have been discussed above:
1.) Lack of templating is not really a problem, just separate the
functions outputting HTML from the rest. The only HTML related problem
left then is that it might seem daunting for designers if they have to
edit the files, they might also cause syntax errors if they
accidentally remove some parenthesis.
2.) For shared hosting PL needs to be implemented as a CGI module to Apache