Networking

Well, that was simple.

I saw this at a local supermarket for €16,99 yesterday. So I thought about buying it to see if I could share the Vonets WiFi bridge among multiple systems.
Amazon listed it for something weird like €9,07 with free delivery. Prime got it here in 22 hours. ;-)

I dunno. Growing up in the ’80s where networking often required blood sacrifices and matched terminators (who remembers the horrid 10-base-2 with that flimsy co-ax?), I kind of imagined this would be hard.

So, LED #5 is the Vonets. LED #1 is the Pi. I basically just plugged everything in and it “just works”. :-)

In an unquantifiable amount of time (I’ll need to buy another multi-output power supply), I can add the older Pi and have ShareFS running. Or maybe the Beagle xM if I can throw together a utility to expose the command line via telnet or something… Now I have SystemDisc, I can create a µSD card for the xM and ditch the horrid slow USB stick I was using before.

As for the rest of the photo – the glowing thing is the OLED fixed to the front of the Pi2. My original Pi is to the left (out of shot). Above the Pi is a five-output USB power supply running the Pi, the Vonets, the HDMI→VGA adaptor, the Vonets fan (I’ve stuck a 1" fan on the Vonets as it runs scary hot otherwise – those little WiFi MIPs boards tend to do that), and the big-ass dual fan you can see there keeping everything cool. It’s 26°C in here (too damn hot!) and with the fan the Pi’s core is around 50°C with a case temperature of 34°C. I’ve set CPUClock to throttle back at 55°C. Yes, I know the Pi can do more but I’d rather be paranoid, thanks…
Oh, and the monitor is the 1440×900 widescreen that I use with the PC (a nice size for watching DVDs). The Pi’s 1280×1024 screen is to the right. Picture here (before my desk got filled with crap including a massive box that is the PC – I should do something about this pigsty…).

At any rate, I was pleased with the plug’n’play networking.

Those supermarket prices are never as good as they first appear, are they?! I often have similar dilemmas with Aldi!

This reminds me, though. A while ago, I bought a Nexx WT3020, which is a pocket sized WiFi router/AP. It allows me to join my Pi via Ethernet to the house WiFi, so RISC OS can see the Internet.

Best bit? The router firmware can be reflashed with OpenWRT (aka LEDE) so you have an uber-configurable Linux box. I have enabled USB stick sharing via Samba. In theory, I could run AccessPlusPython.

Amazing device. Google ‘nexx pocket nas router’. Mine was about £11 from China!

Hardware failures really start to happen when the ambient temperature indoors is in the 40s. Passively cooled devices tend to fail.
Running fan-less is nice but I fully support using them for the longevity of the device. Just don’t forget to clean them!
I’m constantly shuffling cables because there are only a couple of available ports on the router. A Vonets or similar would be nice.

(who remembers the horrid 10-base-2 with that flimsy co-ax?)

My bit of 10-base-2 networking still works! (An old Sun Workstation which only has a coax connector)

But, more importantly, I read that HomePlugs are all effectively 10-base-2. Is that still the case?

who remembers the horrid 10-base-2 with that flimsy co-ax?

I remember it from work and school, but there’s none on my home network.
When I finally networked my house, it was all UTP. Mostly 100-base-t and a
little 10-base-t. And yes, unless a power surge fries a switch, the hardware
just works!

Strange things happen…

I have a Vonets as well. When I connect it to my Pi and start both (Vonets powered by Pi’s USB) the Vonets is ready after RiscOS is. So RiscOS hasn’t seen a network connection and isn’t able to connect. As I want to use it in a PiTop2 with a battery, I don’t want to use a separate power supply for the Vonets.

When I start the Pi with an UTP cable tied to a modem, it recognises the network and start happy. Now I remove the cable and connect the Vonets. When the Vonets is ready, networking is no trouble.

Now my question: can I configure RiscOS in such a way it ‘knows’ a network connection might be on it’s way? That it is prepared when whatever connection is there. The Boot:Configuration:Network:Internet:Interfaces:Configure is now configured via DHCP.

BTW: I don’t mean a small program in the Predesk that stalls the starting up for a while giving the Vonets time to start.

Peter – I can’t speak for your circumstances, but I tend to find that RISC OS is so quick to start up, that things are just generally better if the Pi has a fixed IP address.

Maybe the proposed changes to the network stack may introduce the idea of the IP address being a variable entity – it would be somewhat necessary if we’re ever going to have a hope of supporting WiFi…
…my Android phone. On my home network. Then there’s no WiFi. Then it finds itself at McDonald’s. Then no WiFi. Then back home. One interface, multiple IP addresses. No sweat.

[actually I don’t trust public access points, but you get the idea…]

things are just generally better if the Pi has a fixed IP address.

Well, Rick, it took some puzzling, the article from late Paul Vigay (http://www.riscos.org/networking/riscos.html) made clear how to do it and…

… it works! Thank you. The Vonets still starts too slow, but after a short while RiscOS picks up the settings and connects.

My own version: https://www.heyrick.co.uk/blog/index.php?diary=20130316

I said it before: lots of neccesary and interesting things like software, hardware, developement, configuration etc are pretty concealed in the world of RiscOS. What the search engine doesn’t find, is pretty out of reach…

The search engine, BTW, needed three words: riscos configure network

What the search engine doesn’t find, is pretty out of reach…

True. It’s also a matter of scale. Fifty billion suggestions for doing stuff to Linux versus half a dozen for RISC OS… Which do you think Google will suggest (assuming it actually stops pointing you at StackOverflow for anything remotely “technical”).

It’s hard. Especially when so much is out of date. Poppyfields, how many of those links work? Paul’s site is a preservation area, my own assembler stuff is more 26 bit than 32, and so on. There was a user called, I think, Rebecca? She was maintaining a module database and searching for latest versions. Is that still active?

Seriously, archive.org is possibly our best friend these days…

But what’s the solution? There isn’t one really. The same problem exists for everything. The only difference between us and mainstream is if there’s a problem with RISC OS then maybe three people have written something about it, while if there’s a problem with Windows that’s perhaps three thousand. But anybody who has encountered an esoteric problem will know well that one correct solution is better than a thousand wrong ones or copy-pasted clickbait. Oh look, we’re back where we started, at StackOverflow…

It’s also a matter of scale

Not only that. It’s a matter of concentration as well. All kinds of sites scattered around the globe make it difficult to find one. Some meta sites would help. Sites containing links to a whole lot of other sites.

I have a web site for the computerclub I’m member from: http://users.cuci.nl/pscheele/ACorner/index.html

We used to be a Acorn/RiscOS club and I on the site is a page containing RiscOS links. I’ll start a search and put every working link I can find in it. Need to think about useful catagories, I presume.

I’ll start a search and put every working link I can find in it

Not sure if its any use but I got tired of forgetting about the places I visit so put some links on my own site:

http://www.xsltpro.co.uk/content/links.html

They will probably go out of date at some point because I’m not actively updating it. A more centralized approach might be better, perhaps a page on this site or on some kind of wiki?

Not sure if its any use

It certainly is, thank you, Glen. I added it to my list.

A more centralized approach might be better

Very good idea. Meanwhile I’m building up a large list of old and new sites, but it takes some time… The main thing is to find suitable categories to put the links in…

so put some links on my own site

Yay! I’m finally listed at the top of something. :-)

It might be worth adding a note that the best way to find something on my site is to go to /blog and use the search (as much of the rest of the stuff is outdated and only kept around because I didn’t like how loads of information vanished overnight when Geocities closed down…). The blog is what’s live, but the random nature of blogs makes indexing the information… somewhat less obvious than perhaps it ought to be.

BTW, my search is a simple substring match, nothing fancy like Google or the old Altavista. ;-)

perhaps a page on this site or on some kind of wiki?

Here, if ROOL are agreeable to it.

I think it’s reasonable to imagine that this site and the wiki ought to be around as long as RISC OS 5 is being developed.

The main thing is to find suitable categories to put the links in…

Probably keep it fairly simple as I’d imagine there will be a deal of cross-over between categories…

• Old/Archive (i.e., pre RO5)
  • Software
  • Hardware
• New/Active
  • Software
  • Hardware
  • Emulation
  • Guides
  • Discussion/Fora

It might be worth adding a note that the best way to find something on my site is to go to /blog and use the search

Will do next time I edit the page!

* Old/Archive
* New/Active

Other way around. Things relating to RISC OS 5 should be at the top of the page where the reader will see them first, and things specifically relating to only older versions of RISC OS should be listed below under a heading such as “Legacy (26 bit)”.

Don’t bother with categories. You are listing sites not specific packages. So how would you describe my site? Software? Hardware? Discussion? Guides? It’s a mix of everything.
Better to simply present a list like this:

• Name of site that is a link
  Brief description of the site, noting the important reasons why somebody might want to visit said site.

I think it might be useful to devise a simple protocol here, where a site has an HTML file, such as “rool.html” in the root which can be used as a springboard to items of notable interest on their site. That way, the listing can end with a “More information” link, instead of polluting the Wiki with lots of links to things in each description…

Don’t bother with categories.

When I’m looking for software, hardware, developments etc, I need some grip to prevent endless searches in very long lists. So I plea for some categories with, indeed, a link and a short description.

@ Rick: I don’t fully understand your last statement: do you mean that every site has a rool.html?

There was a user called, I think, Rebecca? She was maintaining a module database and searching for latest versions. Is that still active?

https://www.riscosopen.org/forum/forums/1/topics/2304, by Rebecca Shalfield (probably not to be confused with Rebecca).

Her website seems to be dead, but the GitHub project is still there, so it might be possible to resurrect it (hopefully it was kept up-to-date).

I think the original idea is still sound, it’s just one of those things that needs someone to take charge and produce the specifications & backend software which will be the heart of the system.

When I’m looking for software, hardware, developments etc, I need some grip to prevent endless searches in very long lists. So I plea for some categories with, indeed, a link and a short description.

But who will maintain this list? A list which must be updated by a human is not a reliable solution.

A list which must be updated by a human is not a reliable solution.

I’m afraid it will never be reliable. Not when updated by human, nor when done by computer. You can have a program looking for dead links that removes them from the list, but a temporarily unreachable link might not be dead. Jeffrey, do you have an idea of a not human maintenance?

Meanwhile I’m going to collect as many links as I can find, just because we have to start somewhere… But it will take some time.

I don’t fully understand your last statement: do you mean that every site has a rool.html?

Yes. Because there are two ways this can look:

• OMNIsite
  Information and resources such as this, that, the other, as well as much more stuff.

Or this:

• OMNIsite
  Information and resources such as this, that, the other, as well as much more stuff. More information.

The first way is liable to link fatigue and potentially getting out of date. Anybody who doubts that should browse TVTropes – a giant wiki full of links, many of which are dead.
The second way gives some information without being advertising editorial and points the user to a site specific resource (the rool.html file) that can be laid out as you please. You could link directly to the index of simple sites, but great messes like mine would probably benefit from some sort of categorised index.

not a reliable solution.

There is no reliable solution, only best attempts.

As for the site – I got Rebecca right. Wrong Rebecca, but remembering it at all is pretty good for me…

Jeffrey, do you have an idea of a not human maintenance?

Use a web crawler which populates a searchable database.

There are a number of options by which the crawler could detect things that were worthy of addition to the database – meta tags in HTML pages, special index files like “rool.html”, PackMan package indices, etc. Basically something which the author/publisher of the content can keep updated with minimal effort (have a single “software” page on your site hosting all your software downloads? Stick meta tags in there. Have a blog with articles which cover a variety of different topics? Stick different meta tags on each blog entry). There is a problem here of discoverability for existing content – so a way of manually defining things to be added into the database may be necessary (with appropriate mechanisms in place to make sure that people can’t abuse the system to add fake/inaccurate entries).

Each database item can have a “missing” counter associated with it. If the crawler crawls the page the item used to be on, and the page is missing or the item is missing from the page, the “missing” counter can be incremented. Once the counter hits a certain threshold the item can be removed from the search results (or added to a special “missing” page on the site). If/when the page becomes accessible again the counter can be reset to zero.

So, four things needed:
- a list of sites to fill…
- … a database,
- sites with meta tags like rool.html etc. and
- a crawler to examine the sites.

And in the database some categories to put the links and short descriptions in.

What are we waiting for? :-)

Use a web crawler which populates a searchable database.

Who runs the crawler?

meta tags in HTML pages, special index files like “rool.html”, PackMan package indices, etc.

Indeed, there are many different types of information that may be pertinent – for instance a “how-to” versus a package versus a utility/module versus a resource (updated Messages/Templates or translations, etc).

The crawler would need to be smart enough to tie all of this together in a way that isn’t going to be crazy to try to read. Not to mention that it will need to be clever enough to detect “the latest version” of something to provide links to that, and not various older incarnations. A good example here is Zap – would you like the 26 bit version? Tank’s 32 bit version? Or one of the three (?) that followed ending with what I believe is the most recent – my Zap builds. At least, until somebody else hacks’n’patches and makes another one. ;-)

Stick different meta tags on each blog entry

My blog fails validation for two reasons. Notepad++ ends <img…> tags with a ‘/’ which is correct for XHTML but not HTML4. I think it’s of minor importance so I don’t bother trying to fix it.
And the second thing is the use of two invented tags: <deftitle> and <deftag>. These appear within the HTML document at around line 150. They are not meta tags. The reason for this is that the blog pages are built by:

• Generate the document header (everything up until the email address)
• Insert a marked up document specific to the date being requested (the blog content)
• Generate the document footer (the entirety of the right-column stuff)

When I’m writing a new blog entry (at source level, I write it all by hand), what I see begins with the <deftitle> line and ends before the comment code (it as much as says this in the document source). The rest is automatic.
So on my blog, it’d be looking for “<deftag \”risc os\“>” (quotes escaped) in the document body.

If the crawler crawls the page the item used to be on, and the page is missing or the item is missing from the page

There would need to be a forwarding tag so that the references can be updated if the item in question is moved (without resorting to HTTP redirections, but they will need to be supported as well).

to make sure that people can’t abuse the system

That’s why those of us that remember Web 1.0 have strong and unpleasant feelings about the entire practice of “SEO”. Indeed, I can imagine that a lot of Google’s work is expended in trying to detect those gaming the system, working out how, and coding ways of disabling that.

In our case, it could potentially matter if sites are to be listed in order of some arbitrary metric of “popularity”. Better, perhaps, to sort the database entries and list things alphabetically (says a person whose site is the eigth letter of the alphabet ☺). That way, no favouritism, no gaming (unless I register aardvark.eu! ¹), and ought to be consistent.

¹ Can’t – taken by “Aardvark Language Services”, and you can’t tell me that any company called Aardvark hasn’t done so specifically to be listed first. A Dutch company has registered “aa.eu”² (stands for their initials, but in English it’s a very archaic middle-English word for watercourse/stream). And you can’t have single letter domains, so “a.eu” doesn’t exist. I guess I’m outta luck…

² “aa.fr” is already taken, and “aa.co.uk” is inept (it requires the “www.” prefix) and amusingly is American Airlines, not the expected yellow breakdown truck (“theaa.com”!). Plus, it pastes a suffix to the URL “?locale=en_US”… to the domain www.aa.co.uk. Bloody Yanks…

but the GitHub project is still there, so it might be possible to resurrect it

Oh, okay. Count me out on that. I have zero idea about MongoDB nor do I speak Python, so I can sort of work out what’s going on, but in the same way I can sort of follow Z80 code (but would be useless at writing my own).

BTW, you can (usually?) download a copy of stuff from GitHub by adding “archive/master.zip” to the URL, like: https://github.com/RebeccaShalfield/RISCOSSearchEngine/archive/master.zip
Useful to know with NetSurf that doesn’t have a working download button. ;-)