Planned web site down time
Posted by Andrew Hodgkinson Tue, 15 Mar 2011 13:28:00 GMT
The RISC OS Open web site will be undergoing a few changes in the next day or two. These include an evolution of the existing site design which renders more quickly in NetSurf and replacement of the problematic Wiki engine with one which will hopefully prove more reliable.
It is necessary to take the site offline completely during the upgrade process. Data must be migrated from the old to new Wiki and no changes can be permitted during that time. The web server architecture is being overhauled too. As a result we are planning for an hour or two of downtime in the next couple of days.
Ongoing disruptions possible
There may be ongoing minor disruptions while we iron out any issues that only show up once we’ve gone live with the new system, so don’t be alarmed if you can’t connect. Hopefully the impact will be minimal.
Very old RISC OS browsers are no longer directly supported. While the new site will be usable, it won’t be very pretty! We recommend that you use NetSurf on RISC OS in future.
Users should delay major content changes
If you are planning to make any big changes or updates to site content, in particular in the Wiki, we recommend that you wait until the new site is live before starting that work.
More information
Please get in touch at info@riscosopen.org if you have any questions about the site status; e-mail connectivity will not be interrupted.
Most of this work is now complete.
Things like C/C++ are handled properly, even if C/C++ appears again in the sentence. Some Textile variants used to treat the “+” or “/” as part of a special character sequence.
Some parts of the site may still filter some kinds of markup, for example tables are filtered out in the news comments section here. If you spot any and it annoys you, please contact us.
Things not done:
A wiki with images and proper page history? It’s like you knew exactly what I wanted! ;)
Nice work!
And a list of which pages link to the current page! However you might want to rethink the decision of having the links listed at the bottom of the page – the main PRM index page takes ages to load in Internet Explorer (but Netsurf seems OK, so I’m just going to blame IE’s page layout algorithm).
I just tried it in MSIE 6 under DarWINE. It looked ugly as MSIE 6 isn’t supported and can’t handle the transparent PNGs, apart from anything else, but it rendered almost instantly.
I then tried in MSIE 8 under Win 7 on VMWare. I see what you mean! It just sits there soaking CPU for almost a minute before rendering the page. Way to go Microsoft! I think it just can’t handle a large number of links, or perhaps flowing text; either way, there’s nothing I could do short of removing the page references altogether. Besides, MSIE isn’t supported
:-)– we recommend you use NetSurf or a WebKit derivative like Safari or Chrome.For giggles I tried upgrading to the shiny new MSIE 9 (not being pushed out over Windows Update for some reason, strange; had to install it manually). The big tagline on the download page is “Fast is now beautiful”. Different issues. It shows the page quickly, but as you scroll there a big delays for rendering. Worse, it seems horribly buggy; when I hover my pointer over links, there’s a huge delay, then the whole page reformats to something very wide for no reason at all until I move the pointer away from the link.
Just don’t use MSIE. It’s never worked and it never will.
Microsoft: The More We Change, The More We Stay The Same.
It was all too good to be true…
:-)I’ve identified a serious problem in the bug tracker that causes Ruby processes to get stuck spinning in a tight loop under certain conditions and this interferes badly with the wider hosting environment, especially as more and more such processes begin to accumulate. If you’ve had problems with the ROOL site running very slowly or generating errors instead of serving pages recently, that’s a symptom of this problem.
The fault is triggered by attempts to view particular files in the Subversion repository. Something goes wrong inside a piece of code that handles syntax colouring. This code has not been changed in the recent web site upgrade so the failure is currently a mystery and probably points to bugs or incompatibilities in supporting pieces of software, such as shared bits of library code, or possibly even the Ruby interpreter itself – we are using a faster, lighter weight Ruby interpreter for the updated ROOL site compared with the one that we ran before.
As a temporary measure I have disabled Subversion browsing. RISC OS code is in CVS and that’s unaffected – only the web site sources in Subversion cannot be viewed through the web interface. You can still download tarballs and look at Changeset information.
If you try to browse Subversion, you’ll be asked to log into Hub if you haven’t already, then additionally asked to log into the bug tracker itself. This is something a public user would never normally have to do. No public login credentials are available. The whole point is that access to this mechanism is temporarily shut off in order to basically save the bug tracker from itself!
I don’t know when this will be fixed but hopefully not many people will care too much anyway, since the overwhelming majority of interest is in the RISC OS sources in CVS.
Hopefully this is now fixed so source browsing has been turned back on again for the Subversion tree.
What about that down time, then? “Failed to connect to post 80: Connection refused”. Was that planned?
There were even more infinite loop bugs in the YAML syntax library so I just got rid of it in the end.
Obviously there will be unscheduled down time occasionally; there’s no such thing as bug free software or perfectly reliable hardware.