Okay, replying to myself and topposting... *hands some spears*

I made a test and simply replaced the mimetype with "image/gif", so the 
"image" is broken but the browser still thinks it's an image and 
hopefully behaves normal.
My Opera 9.23 sends a new request after 5 seconds, so from my side it 
seems to work with binary based mimetypes.


freenetwork at web.de wrote:
> Sorry, I don't have a capable HTTP framework to return binary data. I'm 
> limited to console commands...
>
> PS: "Refresh"s upon a 301 redirect don't seem to work here.
>
> Matthew Toseland wrote:
>   
>> The question is, does it work with images? When I tried it (some time ago), 
>> it 
>> didn't work with inline images.. well I think it didn't, could you check 
>> please? multipart/replace *might* work with images. iframes would definitely 
>> work with images but require that we know the dimensions of the images in 
>> advance.
>>
>> On Thursday 27 March 2008 17:33, freenetwork at web.de wrote:
>>   
>>     
>>> Matthew Toseland wrote:
>>>     
>>>       
>>>> On Thursday 27 March 2008 01:16, you wrote:
>>>>   
>>>>       
>>>>         
>>>>> * Michael Rogers <m.rogers at cs.ucl.ac.uk> [2008-03-26 09:36:32]:
>>>>>
>>>>>     
>>>>>         
>>>>>           
>>>>>> On Mar 25 2008, Matthew Toseland wrote:
>>>>>>       
>>>>>>           
>>>>>>             
>>>>>>> Anyone got any better ideas?
>>>>>>>         
>>>>>>>             
>>>>>>>               
>>>>>> Sorry if this would be impossible, I don't know anything about fproxy's 
>>>>>> internals, but when a key is requested, would it be possible to display 
>>>>>>           
>>>>>>             
>> a 
>>   
>>     
>>>>>> "please wait" page with a "cancel" button and/or a link to the main 
>>>>>>           
>>>>>>             
>> fproxy 
>>   
>>     
>>>>>> page, then redirect to the content once it's been fetched?
>>>>>>       
>>>>>>           
>>>>>>             
>>>>> Sure it's possible... but not a realistic option. If the browser happens
>>>>> to have reached its maximum amount of connections, I bet it won't obey
>>>>> the "refresh" header we would set on the page we display, effectively
>>>>> "breaking" the browsing.
>>>>>     
>>>>>         
>>>>>           
>>>> Why would it be unrealistic? It wouldn't reach its maximum number of 
>>>> connections in the first place because connections wouldn't be blocked.
>>>>
>>>>   
>>>>       
>>>>         
>>> IMHO the only sane solution.
>>> I made some tests (Opera 9.26) with a refresh statement in the HTTP 
>>> Response header and they were cheering.
>>>
>>> - run " nc 127.0.0.1 -p 5555 -L"
>>> - point browser to http://127.0.0.1:5555/
>>> - request can be seen in console
>>> - send back this:
>>> "
>>> HTTP/1.1 200 Found
>>> Refresh: 5
>>> Content-Type: text/plain
>>> Content-Length: 10
>>>
>>> Welcome...
>>> "
>>> - browser displays the stuff
>>> - after 5 seconds the browser again requests the URL
>>> - maybe this time the final content is returned, without the "Refresh"
>>>
>>> PRO:
>>> - it does not use any HTML meta header, so we can send back "untampered" 
>>> content that get requestes after a certain amount
>>> - therefore it might work on other content types as well, like images, etc.
>>> - for every delivery a Content-Type can be set, so you can have e.g. 5 
>>> HTML-pages saying "please wait" and after the file has been found it can 
>>> be e.g. an PDF
>>> - if we know an image is requested, send an placeholder image (e.g. gif) 
>>> and let the browser retry later. then, after the final image has been 
>>> found, send the real image (e.g. jpg)
>>> - the connection is like, never, open, and therefore not accounted from 
>>> the browser, making a connection limit meaningless
>>>
>>> DUNNO:
>>> - do other browsers support this?
>>> - does it work only on HTML or other file types as well?
>>>
>>> CONS:
>>> - nuttin ;)
>>>
>>>
>>> IMHO every tweaking with browser settings is fixing the symptons and is 
>>> very browser-specific.
>>> If this works for every (supported) browser, we can ease everything :)
>>>
>>>     
>>>       
>>>> The tricky bit is images - if they can't be loaded immediately, we might 
>>>>       
>>>>         
>> have 
>>   
>>     
>>>> to convert them to iframes or something. Which means we'd have to know how 
>>>> big they are. But if there are no dimensions specified? 
>>>> multipart/x-mixed-replace maybe, but some browsers may not support it?
>>>>   
>>>>       
>>>>         
>>>>>> It seems to me 
>>>>>> that people are opening lots of tabs, nothing is loading so they suspect 
>>>>>> Freenet has crashed, they try to open the stats page and when that fails 
>>>>>>       
>>>>>>           
>>>>>>             
>>>> it 
>>>>   
>>>>       
>>>>         
>>>>>> confirms their suspicion. So rather than changing Firefox, maybe we just 
>>>>>> need to give the user some feedback that Freenet is fetching the content 
>>>>>> and it hasn't crashed?
>>>>>>       
>>>>>>           
>>>>>>             
>>>>> I don't think it's why people open loads of tabs... Freenet's latency
>>>>> sucks hence "parallelizing" makes sense
>>>>>         
>>>>> ------------------------------------------------------------------------
>>>>>
>>>>> _______________________________________________
>>>>> Devl mailing list
>>>>> Devl at freenetproject.org
>>>>> http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
>>>>>           
> _______________________________________________
> Devl mailing list
> Devl at freenetproject.org
> http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
>
>   

Reply via email to