[Typo3-dev] Typo3 4.0 wishlist ;) <rant>

[Johannes Reichardt]

Hi Robert,

> 
> Well, if you have a look at how caching works internally you'd see that we
> have two options: Showing a blank page or a "Pages being generated"
> message. The point is, that we have to show _something_ if a page is
> accessed which already has been started to render but is not finished yet. 
> 
> BTW: You can configure the "Page is being generated" message, should be easy
> setting it to an empty string.

Obviously that is not the point. The way it goes with searchengines 
(which really have to be respected today) is the problem. Both ways are 
not well - maybe the service unavailable method makes sense, i wonder 
how long that "generating" will take normally and if someone cant just 
wait 5 or 10 seconds these days without seeing something instead.


> Changelog RealUrl:
> 
> 2004-06-10  Kasper Skårhøj
>    * For 404 handling, realurl will now use the default TYPO# pageNotFound
> handling.
> 
> Which means: You can configure it to send a real 404 message.

Yes, i have read about it also. But how :) - i use  that suggested 
technique (will not go into detail here) but it is still giving me a 
very hard time to find out how. Also - and thats the point why i put it 
on the dev-list - these problems occur on many installations and people 
even dont know about it, resulting worse ranking etc. SE friendly pages 
are extremly important today - (search for "postvarset error" in google 
gives 34.000 pages ;)

> 
> 
>>Yet another problem is having realUrl links that are only "one way
>>tickets" - means that they have to be in the cache to be valid,
>>otherwise they do throw an 200, a 404, the startpage, the next directory
>>level, nothing or just something. That is extremly bad for "link
>>transparency" as well.
> 
> 
> So, what do you suggest? It's the same with indexed_search: A page has to be
> in the cache so your can search it.

In case of the linking a complex algorithm could try to look up the page 
if its not in the cache (which would have bad performance for that time 
but at least it would work uncached) - i will think about it, but if its 
really dependent to some cache there should be a built in button that 
regenerates the cache.


>>Maybe there is some sort of "typo3 light" in the future with a
>>bullet-proof built in realurl/static_html implementation and
>>compatibility and built-in options overhead in general?
> 
> 
> The problem is: You want it this way, I want it differently. That's why it
> is configurable.

Yes thats true - that makes it so hard to complain. Actually i can 
configure almost everything, on the other side just the feature-richness 
makes a "normal" php programmer go crazy. I could have "simulated" the 
realurl behaviour within a few hours with my own scripts and that whould 
be highly customized and i would be able to fully understand it.

You know what i mean? I don´t want to blame typo3 for that, but at least 
these things about realurl are too important to be treated that way to 
my opionion.


> <backrant>
>    Well, I think that most of this is solved and that your suggestions come 
>    a bit late for 4.0. They could have been good suggestions for the Zap The 
>    Gremlins.
>    Of course I can't require you to know all technical background of the 
>    TYPO3 caching mechanism, but if you really suggest changing it, I ask you
>    politely to look at the implementation and propose a realistic 
>    alternative. It might look easy at the first glance, but it's quite 
>    complex ...

I know these things are complex and unfortunatly i am not good enough to 
  make better suggestions. But at least that "there are things going on 
you can´t really control due to caches or something" seem very dangerous 
to me in the long run. Maybe some more straight forward, transparent 
implementations (even if they´re slower) or just some more indexes or 
just 3rd party solutions, advices or presets would make more sense.

I still dream of a typo3 light that ouputs just nothing at first without 
templates and hardly used options etc... when i start to define some 
"codeparts" it will output them, but no more newline than i put into it :)

- Johannes