Re: mathopd - what else? :)

On Sun, 7 Jul 2002, Thomas E. Horner wrote:

>
>
> dear folks!
>
>
> i searched a lot.
> the cause is simple and may seem unlogic to some people:
>
> i try to never use a product with a _very_ wide spread.
> i try to never use a product that has an overloaded feature-set.
>
> that's why i would never use apache.
> and the worm discovered a short time ago tells me i'm right.
>
> i found mathopd.
> and it is the best webserver i've ever worked with.
> i just love it.
> the only bad thing is that some of the best features have
> to be read out of the code (safe-shutdown, rotating logfiles, ...)

Right. This will be in the 1.4 docs. Sorry for not finishing this sooner
but I'm working more or less 7 days a week until august 1st, so do not
expect anything until that date.

> i'm a coder (at least i think :)
>
> phpstub was buggy.
> i changed a few things.
>
> now it can hang only for a defined time and then kills itself
> and php (set to 5 min)
> it now can never happen that phpstub uses 100% cpu (which
> is what it did in some cases before the change)

Hm, that sure sounds like a bug (the 100% thingy). Is there a sure way to
reproduce this?

> i added accept-language, accept and accept-encoding to mathopd
> cause i need to have this info in php, which is what i mainly build
> my sites with.

Right. The problem is at the moment that mathopd only passes some HTTP
headers to the application, rather than all of them. I will try to fix
this in future versions. Anyway, mathopd was not really designed to work
with anything other than static HTML files and images.

> i'm missing 1 feature and found 2 terrible bugs in mathopd (all versions)
>
> feature missing:
> 	i'm selling webspace and so i need to be able to count not only the
> 	traffic of static files but also of cgi's and other external processes
> (php, ...)
> 	it should not be too hard to pipe the output through mathopd, is it?
> 	then the correct return-status (default 200 or the value php returns, if any)
> 	and also the byte-count of the output-data could be correctly logged.

Counting bytes sent through cgi scripts can be done in phpstub. I believe
there is some logging code left in there that I might be able to recover.

> bug:
> 	go to http://www.t-horner.com/secure
> 	user: mathopd pwd: mathopd
> 	notice the missing (!) trailing slash
> 	if it comes up (ie if using internet explorer) then re-query the page
> after login
> 	after removing the trailing slash
> 	scroll down to 'environment'
> 	PATH_INFO is set incorrectly, it should contain an trailing slash (BUG)
>
> 	go to http://www.t-horner.com/secure/xxx/yyy/
> 	which doesn't exist
> 	again the page (/secure/index.php) comes up although an error-404 should
> 	be the correct answer. (BUG)

Ok I don't have a browser or the source handy where I am right now but I
will try this ASAP.

Cheers
Michiel