[www.sljfaq.org] various things

[Ben Bullock]

I've recently been doing some work on the book page to deal with bad
inputs. As a byproduct of the work I accidentally was sending the book
page without the bottom part of the HTML which contains the Google
Analytics code among other things. That means the Google Analytics has
not been working which I just noticed today when I noticed I had no
visitors.

As a result of the exasperation of not being able to use the nutty
HTML validators available on the web due to the false positives they
produce, I spent today making an HTML validator of my own. Anyway now
at last I have a workable validator I've been back to not only the
book page but the sci.lang.japan FAQ pages and validated them, which
meant about two or three hours of fixing broken HTML in Emacs. All of
the pages should be valid in terms of not having stray HTML tags and
so on in them. I'll be doing more work on this validator so that all
the CGI pages are automatically tested and this kind of accident with
the books page can be automatically detected. Basically it would be as
easy as noticing that the page had no </html> tag at the end of it.

I've altered the dates page

http://www.sljfaq.org/cgi/date.cgi

to produce the current year in the Japanese format by default. That is
because the main use people seem to have for the page is to get the
Japanese year for 2015, so since it's calculating this year anyway as
part of the execution it may as well have that as the default output,
and hopefully be a little more useful to people.

I've moved a page from www.lemoda.net to
www.sljfaq.org/lemoda-numbers/. This was the last CGI program left on
Lemoda.net and I've removed it so that I can switch Lemoda.net to
being an HTML-only site. That means one fewer headache in terms of
ongoing administration of these web things and frees more time to do
the sljfaq.org site.

Further to all of the above I've been doing various jobs with the cgi
programs on sljfaq.org related to errors occurring with some inputs,
which I won't detail here.

[Ben Bullock]

I was so pleased with this validator that I spent the rest of today
running it over all the HTML for individual pages. Some of the HTML
bugs are twenty years old! Since my validator doesn't catch things
like <p><p> it missed a few bugs which were found by running over the
individual pages. Now after about an hour and a half more of emacs and
a Perl script to do batch editing of the broken table entries, all but
one of the pages are OK by the validator. The final exception is the
"encodings" page which contains a lot of HTML which I didn't write, so
it uses features I didn't use and the errors are harder to remove. So
only the encodings page is not validated with the new method.

Anyway I am pretty pleased after twenty years that I can finally
validate the HTML successfully.

[Ben Bullock]

Further to the previous post, I've been using the HTML validator extensively to catch errors. I've overhauled most of kanji.sljfaq.org and most of the CGI programs at http://www.sljfaq.org/cgi/ so they produce better HTML. I also found an online validator which is quite useful at http://www.onlinewebcheck.com/. This caught a lot of problems which my system was too primitive to notice. I still haven't finished validating some third-party code like the encoding page, but already I feel much better about having a way to test the HTML (at last).