Netsurf 3.3 disc cache problems

Since upgrading to Netsurf 3.3 it’s been occasionally complaining back to me that my disc cache isn’t big enough, and telling me to either make it larger or turn it off by setting it to zero.

However, the disc cache is already set to 4095 MB and it won’t let me make it any larger.

OTOH, if I set it to zero I get all sorts of weirdos happening – until suddenly the weirdos go away and…

..it complains that my disc cache isn’t big enough, and when I look at the configuration it’s gone back to 4095 MB.

 
WTF?

Mine keeps saying that it’s too slow and has been disabled – but I can’t see how as it’s an i7 pc with very fast disk access more than 150mb/sec. Also it gets stuc loading google in an endless loop, and an error box appears forever and I have to reset as I can’t clear it (looking for save file).

Also it gets stuck loading google in an endless loop, and an error box appears forever and I have to reset as I can’t clear it

See my posting here

But back to the subject…

What should I do about the Netsurf disc cache?

I tried Memphis.. seems to have not come back since or you can turn cache off.

If you can’t find a better solution, you can remove the disc cache altogether. By default it gets placed in Cache. If you move it somewhere else where it will not be seen, NetSurf will operate without as far as I can tell.

Alternatively, ask on the NetSurf mailing list for advice, or report the issue to the developers: they cannot improve things they have not had drawn to their attention.

If you can’t find a better solution, you can remove the disc cache altogether. By default it gets placed in Cache. If you move it somewhere else where it will not be seen, NetSurf will operate without as far as I can tell.

A better solution would just be to turn the cache off in NetSurf’s configuration dialogue, by setting its size to zero. Note that the problem is the disc cache, not the memory cache.

If this doesn’t work, please report the fact by raising a ticket on NetSurf’s bug tracker.

Alternatively, ask on the NetSurf mailing list for advice, or report the issue to the developers: they cannot improve things they have not had drawn to their attention.

I think all the questions here, and in other recent threads (here and on csam) have been discussed at length on the NetSurf Users mailing list prior to the release of NetSurf 3.3. The list should really be required reading for anyone using NetSurf: certainly there’s been some detailed discussion as to the reasons that people are seeing the cache “error” message, and what (if anything) they can do about it given the limitations of RISC OS.

We need to remember that most of the core NetSurf developers don’t use RISC OS. While a lot of work has been done recently to improve the disc cache performance on our systems, it’s being done “blind” and feedback from users of the RISC OS test builds is required (as has been explained on the mailing list). The developers are also dependent on bug reports (via the bug tracker) from RISC OS users, as otherwise problems specific to our platform will very likely go unnoticed or forgotten.

It’s worth stressing again why bugs need to be reported via the tracker, as I know some people dislike it. Tickets on the tracker are visible and their progress can be followed; they won’t get forgotten, or fall through the cracks. Posts here, or even to the NetSurf mailing list, are far more “transient”: it’s easy to forget them after a few weeks, and then problems don’t get fixed.

I think all the questions here, and in other recent threads (here and on csam) have been discussed at length on the NetSurf Users mailing list prior to the release of NetSurf 3.3. The list should really be required reading for anyone using NetSurf:

I don’t think the NetSurf developers help with that bit as the only mention of the mailing list I’ve found is tucked away on the NetSurf > documentation > User Info page well down the page under “Feature Requests”¹
Wouldn’t putting a link from the home page be a bit more useful / obvious?

¹ Rather Douglas Adams planning notification-ish really.

Wouldn’t putting a link from the home page be a bit more useful / obvious?

Is http://www.netsurf-browser.org/contact/ obvious enough? It’s linked to from every page of the website…

Is http://www.netsurf-browser.org/contact/ obvious enough?

Since you ask – No. Oh, the text is visible enough, the message conveyed is not “here is a mail list” though.

It’s linked to from every page of the website…

And gives the wrong impression. People the world over are used to the idea that clicking on a “contact” link will fire up their mail client and pre-populate the To field with the mail address of the mailbox the web author wants you to use as a “contact”

Now if you saw a link that said “mail list” you’d be thinking ‘ah, that’s where I click to get to the mail list information’ rather than ’that’s where I click to get a contact the author mail message shell’.

People the world over are used to the idea that clicking on a “contact” link will fire up their mail client and pre-populate the To field with the mail address of the mailbox the web author wants you to use as a “contact”

Remember, NetSurf is not a single person effort, so it is more likely that the Contact link would open a page with either a list of contacts or an annoying on-screen form to fill in. Examples…

• https://www.riscosopen.org/content/contact – from link at top of page
• http://www.bbc.co.uk/contact – from link at bottom of page
• http://www.nhs.uk/aboutNHSChoices/Pages/ContactUs.aspx – from link at top of page (it is a form)
• http://www.amazon.co.uk/gp/help/customer/display.html/ – from “Help” link at bottom of page, then at the bottom of the “Need more help?” options (and it is a form)
• https://www.google.fr/intl/en/contact/ – from “Contact” link at the bottom of the “About” page. Good luck finding an actual email address, though…
  
  

Either way, while there is a sort of logic to having the Mailing List information under Contact, I would agree that if I was tasked with looking for the NetSurf mailing list, I would only find it there after having first tried every other link (except Download); assuming I didn’t just “Search” for mailing list and see Google’s paid adverts for people selling email addresses and padded envelopes…
…though this presupposes that I had advance knowledge that there was a mailing list. I wasn’t aware of it until Matthew’s post five hours ago; because… well… if I didn’t feel a need to contact one of the developers, why would I look at the Contact link?

Note that the problem is the disc cache, not the memory cache.

No, the problem is NetSurf arbitrarily deciding that the caching is “too slow”.

The chart here indicates that a Pi Model B+ can access the “harddisc” at about 90mbit/sec, the harddisc being a class 10 SD card. 90mbit/sec is approx. 12mbyte/sec. Let’s round it to 10mbyte/sec for ease of calculations and to account for protocol overheads. If that’s a class 10 card and I’m using a class 4 card, then that will imply that my card can manage 4 megabyte/sec (indeed, that is what “Class 4” means). Even if this is burst speed, not sustained, it is certainly not a slouch.

My Internet runs at 2megabit, which means it can achieve around 265kilobyte/sec flat out (as measured by mIRC and Firefox when downloading animé). The Pi does not manage that. I don’t know if it is the Vonets adaptor, the weak signal, or the Pi itself, what I do know is that the Pi manages around 90-140kilobytes/sec (as measured by NetSurf). To put this into comparison, the internal IDE harddisc in my A5000 could benchmark that sort of speed (the Connor CP3044 rates as 1.5MB/sec reading from media), and that was when a 42MB(base 10) harddisc was sufficient. Nowadays, most of the videos I watch on YouTube are larger!

Anyway – if the data comes in at ~140kilobytes/sec and the SD card can accept data at 4megabytes/sec…let’s be savage to the RISC OS filesystem and say that it is only 25% efficient and tops out at 1megabyte/sec on my setup. That is still over five times the speed of the incoming data (leaving plenty of time for processing); which surely implies that a disc cache would be beneficial and that NetSurf’s method of calculating how fast (or slow) it is . . . is faulty.

Either way, while there is a sort of logic to having the Mailing List information under Contact

I work in tech support and I can assure you that if it’s possible to be confused then the users will be.
The name of the game is making things as clear as humanly possible, mostly by doing things that fit with the way that the users think rather than the way that techies think.

…though this presupposes that I had advance knowledge that there was a mailing list. I wasn’t aware of it until Matthew’s post five hours ago; because… well… if I didn’t feel a need to contact one of the developers, why would I look at the Contact link?

That was rather my point. The rest of your post provides emphasis.

Since you ask – No.

I’m sure that, as with everything else NetSurf related, if you got in touch with the volunteers who develop it in their spare time and suggested a sensible improvement to the site, they’d be happy to discuss it. They might ask you to make the changes and send them in¹, but that’s only fair given what you’re paying for the service.

¹ Like everything else, the website source is held in the NetSurf Git repository. You can download a copy, make changes and submit a patch for consideration.

No, the problem is NetSurf arbitrarily deciding that the caching is “too slow”.

No, the problem of the error message is caused by the disc cache being enabled (and NetSurf thinking the disc is too slow). Disable the cache, and the message goes away.

As for your figures about access speeds, there’s been a lot of discussion on the mailing list about the issue, the way that NetSurf’s cache works, and how the developers could improve it (it works fine on non-RISC OS platforms, apparently). Changes have already been made in the hope that they might help.

AIUI, current versions dump a lot of information about the cache usage, and why it thought the disc was too slow, to the log file when the browser is quit. It has been requested that users send this information in to the developers so that they can see what’s going on in the real world. Lengthy rants here aren’t going to help; running NetSurf on your system as described here and sending in the cache performance output will be useful.

Again, if you’re using NetSurf – especially the test builds – you really need to be reading the users mailing list so that you’re aware of work-in-progress like this. If you’re using a test build, you’re effectively an alpha- or beta-tester (depending on where things are in the release cycle).

Suggestion to help the netsurf volunteers: Change all links saying “Contact” to “Contact/MailList”
Ideally it would be either an extra link saying “Mailing List” or “Contact/Mailing List” but that would be too long for the links bar.
I know some people don’t like Forums but I think something like NetSurf would be more effective using a forum rather than a mailing list.
I can’t see an online archive of the mailing list!

I can’t see an online archive of the mailing list

Oh, if you follow the links far enough it exists. The link is near the top of that page.

BTW. Remarkable co-incidence but I came across this while testing whether BT were going to stop p***ing around with my adsl link.

Essentially it points out how bad it can be for designers to not design things as users expect. Changing small details in this instance increased the usability of the specific feature by up to 22.4%

I wonder what the small change Chris suggests would do to the usability of the NetSurf site?

¹ I could walk down to the end of the road and kill the guy fiddling in the green box, but apparently that’s frowned on.

With the Archive section having a Heading of: “About netsurf-users” had made me think it was about the users of netsurf and I could skip it and move on! Rewording that heading would be a good idea.
I can’t see a way of searching the Archives:-(

I can’t see a way of searching the Archives:-(

in a google search box specify the search term thus:

Theoretically that should search for the term “disc cache” and limit the search to only looking for references on the site listed.

Re. http://www.bbc.co.uk/news/magazine-31602745

BTW. Remarkable co-incidence but I came across this while testing whether BT were going to stop p***ing around with my adsl link.

Are you sure it is only his link he is messing with? For the past five minutes, Firefox said that BBC stopped responding to the request; though I have it now and – no, it looks nothing like a hamburger if you ask me. It looks like… well… a drain grill.

Re. mailing list archives

You have to read every single message one by one? What is this? The nineties?
<facepalm>

The RPi 1 and Pandaboard ES seem to really struggle using the Netsurf disc cache whereas the RPi 2 and IGEPv5 seem fine.

no, it looks nothing like a hamburger if you ask me

Me either. I see a pattern like that and I see it as alternating network switch and cable management¹

Are you sure it is only his link he is messing with? For the past five minutes, Firefox said that BBC stopped responding to the request

Some requests – depended on which story you were trying to look at.

You have to read every single message one by one?

Well you could download the gzipped text and then read that after extracting the content.
That would be a month at a time.

I have a nice perl script on a linux box at work that would search all the .gz files for specific strings and return the results line by line. Not much use round here though.

¹ But I’m the one that can draw the network topology for all the various elements of the LAN (two sites) the extensions via VPN links, firewall DMZ’s etc from memory and not remember family birthdays.

Let’s round it to 10mbyte/sec for ease of calculations and to account for protocol overheads.

No, that doesn’t account for protocol overheads at all. The disc cache has to store files. You must have noticed that you can’t stream an average of 10 MB/s into files that have to be opened, written to and then closed, when you include the time to open and close them.

Re. mailing list archives

You have to read every single message one by one? What is this? The nineties?

There are many web interfaces to the public mailing list archives. I found these for you with Google:

https://www.mail-archive.com/netsurf-users@netsurf-browser.org/
http://blog.gmane.org/gmane.comp.web.netsurf.user/

There are lots of others too.