[Opera-Linux] How to reload faster?
Daniel Eckl
daniel.eckl at gmx.de
Sat Jun 9 21:46:09 UTC 2007
Hmmmmm... my hitcounter gif image has been loaded from 64.154.81.197 (no
DNS name for this IP) which differs from your dump
(farnsworth.tigertech.net has address 64.62.209.18).
Don't know what it does there or if this is related to the history back
action or some other application / open website on your machine did ask
for this server...
Hmmm what I could think of is: I use excessive as banner blocking.
Perhaps this is from an ad banner haveing the "no-cache" pragmas as well
and opera reloads it every time you go back to digg.com. If the server
where the banner resides is slow, it could cause some delay.
Best,
Daniel
Larry Alkoff schrieb:
> Thanks very much for your informative response Daniel.
>
> My machine uses a 950 mhz Athlon so that would account for some of my
> slowness. I'll be getting a much faster machine pretty soon and will
> check the speeds of a reload.
>
> With \tcpdump -v I get the following for the reload:
>
> 16:00:08.353743 IP (tos 0x0, ttl 64, id 16772, offset 0, flags [DF], proto:
> TCP (6), length: 52) kinda.56633 > farnsworth.tigertech.net.www: F, cksum
> 0xde9a (correct), 513:513(0) ack 501 win 216 <nop,nop,timestamp 405712733
> 830627882>
>
> 16:00:09.037755 IP (tos 0x0, ttl 64, id 22699, offset 0, flags [DF], proto:
> TCP (6), length: 52) kinda.56634 > farnsworth.tigertech.net.www: F, cksum
> 0xb7a9 (correct), 518:518(0) ack 12356 win 997 <nop,nop,timestamp 405712904
> 830627986>
>
> Does this seem right to you? About the same as you got? It seems to
> indicate about a 2 second lag plus cpu time getting the cache information.
>
> There is so much other stuff coming in it's difficult to focus what's
> coming back from the reload.
>
> I agree it would be nice if option 4 (cache in memory) would be implemented.
>
> Larry
>
>
> Daniel Eckl wrote:
>
>> I now tried this. Going back to digg.com takes about 2 seconds for me.
>> I'm running on a Dual Core 1,8 Ghz. If your CPU is slower, it might take
>> longer to render the content, so 5 to 10 seconds sounds possible.
>>
>> I traced my network traffic with wireshark and I I found out that one
>> thing is being transfered when going back: A hit counter image from
>> another server of digg.com. This image is being transfered with these
>> HTTP Response Headers:
>> Date: Sat, 09 Jun 2007 09:43:01 GMT
>> Cache-Control: no-cache, private, must-revalidate
>> Pragma: no-cache
>> Expires: Sat, 09 Jun 2007 09:43:02 GMT
>> (as you can see, the expiration date is on second after the response)
>>
>> So it's wanted that a browser fetches this image every time.
>> Now I think that a history back action is thought to fetch EVERYTHING
>> from the cache to be able to show the exact state you saw when you
>> visited it. But now there's a dilemma: Opera is not allowed to cache
>> this image when it first gets it. But when you go back in history, opera
>> is supposed to show it to you...
>>
>> So this would give three options:
>> 1: Fetch it again from the source URL
>> 2: Not show this image at all when going back in history
>> 3: Always cache everything opera loads, even when the server does not
>> allow to.
>> 4: Cache these kind of content just in memory.
>>
>> At the moment opera seems to use option 1.
>> Option 2 could lead to lots of angry support queries if visible content
>> it missing because of that. Not wanted.
>> For privacy and security reasons, option 3 is not a real option. If
>> server sais "don't cache", then opera is supposed to exactly do so. The
>> content marked this way might contain private information that no
>> attacker should be able to read from your harddisk.
>>
>> But I think option 4 doesn't sound that bad. Caching this content in
>> memory only should be rather safe. Well you can be attacked with a
>> rootkit which is able to read from your RAM (well I'm not a security
>> expert but this sounds plausible to me) but if you have such a problem,
>> then all your traffic and all your keystrokes are compromised as well
>> anyway.
>>
>> I think this would be a good option to handle this kind of content.
>> Is there any opera dev who can comment on this? Is there any drawback I
>> didn't think of? Is this doable or even planned already?
>>
>> Best,
>> Daniel
>>
>> Larry Alkoff schrieb:
>>
>>> Daniel Eckl wrote:
>>>
>>>
>>>> Hmmmmm okay, this way we excluded wrong version as well as wrong settings...
>>>>
>>>> I ran out of ideas now, but perhaps not the opera devs. Could you give
>>>> us some easily accessable example URLs, where the devs can try to
>>>> reproduce that?
>>>>
>>>> Best,
>>>> Daniel
>>>>
>>>>
>>> Open http://digg.com/ in Opera.
>>> Click on an article, then use the back button or z to return to the digg
>>> first page. It takes me 5-10 seconds and seems like it's reloading from
>>> the internet, not from cache.
>>>
>>> Larry
>>>
>>>
>>>
>
>
>
>
More information about the Opera-Linux
mailing list