Heading For The Sun

Yesterday, we hosted our final Sun tech. demo for this semester and for my term as Campus Ambassador. As mentioned previously, we invited Andrew Haley from RedHat to speak about OpenJDK and the work that is being done on making this available in GNU/Linux distributions. Andrew gave a couple of excellent talks: the main one was on the general topic of Free Java and OpenJDK, a potted history of where we were and where we are now, and followed this with a shortened version of the talk he gave at FOSDEM on the TCK (Technology Compatability Kit) for Java.

Unlike our previous demos, this was very much a talk on social and ethical issues rather than a technical one. While many of our students have used Java as a language and Sun’s JDK as a tool in order to write programs, there is generally little consideration of how it can be used in the real world. Andrew’s talk explained this well, clearly illustrating the problems with the previous way Sun’s JDK was being distributed and the problems this raised for those who wanted to develop and distribute Java programs as a part of operating systems like Fedora. From the questions afterwards, it was clear that many were by the various versions of the JDK coming from Sun and Andrew helped to clarify these issues.

As usual, I’ll put recordings of the talks online. As my period as CA comes to a close, I’d like to take this opportunity to thank all of those at Sun who have worked with me over the past six months, and all those here at the university who have attended our tech. demos and helped out where necessary. Your support is much appreciated, and I look forward to welcoming our future campus ambassador when he or she is appointed in the imminent future.

This week is going to be quite quiet by comparison with the last two. Last week, in addition to going to see rms speak, we hosted our Sun SPOT demo with Bernard Horan. Dr Gordon Manson, who I’ve been working with on two modules recently (COM162: Object-Oriented Design and Programming with Java and COM3170: Concurrent Systems), was very involved in this, as he’s been working with SPOTs for years. They also recently formed a part of the teaching material within the Computer Science department, with our COM162 students’ first assignment being to write a SPOT midlet using the Sun SPOT emulator. They had to use the LEDs, switches and accelerometer in some interesting way (Gordon’s trying out a method of ‘let them be creative and thus stop plagiarism’). At present, they are working on a robot controller which could be used with the SPOT-based implementation Gordon has been working on.

With all this in mind, a Sun SPOT demo as part of the CA scheme seemed essential. I would have preferred we’d been able to sort this out sooner, but this is a case with a lot of the CA program this year; things just didn’t start early enough. Hopefully, next year’s ambassador will get a much earlier start on things and not miss the entire first semester. Our students are only with us for 24 meagre weeks, which is barely enough time to do anything. Anyway, we eventually arranged, via Kim, for Bernard to come up and speak on SPOTs and show a few demos. Unlike the previous two demos, where the speaker(s) arrived just before the talk, we met up with Bernard around 12 o’clock and went for lunch first. This gave Gordon plenty of opportunity to discuss his ideas with Bernard and was probably the more useful part of the day.

We were worried that the lecture theatre we had booked was going to be too small. It could only hold about 120 people, and we were expecting a big turnout, including most of the 80-odd students from COM162. But in the end, only 14 people turned up. I can think of several reasons for this. None of these were COM162 students, so I guess the delay between them working with SPOTs and the demo meant they lost interest. I believe some of them chose to instead use the time to work on the assignment, which is fair enough. Note that this was on a Wednesday afternoon (as is our next demo) specifically so there were no lecture clashes.

Our small audience was a good audience nonetheless, and they seemed interest in what Bernard had to say. Gordon also used the opportunity to demonstrate his work on the SPOT-controlled robot (converted from being a TINI robot) which he only just managed to get working, and Tom Gummery demonstrated the third-year project work he’s done with Genix chips to which Gordon would like to port Squawk, the Sun SPOT Java-in-Java VM.

We had a similar small but dedicated audience for our talk the week before. This was on Xvm and so the small size was more expected. Operating systems are not taught with the Computer Science syllabus at Sheffield, so there is no real opportunity to connect OpenSolaris and its technologies with teaching material, as there may be in other universities. Instead, our talk was attended by the subset of UNIX users (and mainly GNU/Linux users) who play with some of the more advanced stuff like virtualisation. Again, I think the talk went well and there were certainly plenty of interesting questions. Thanks to Gary Pennington for travelling up to speak on this.

Our final demo will take place on the 14th (the semester ending, bar exams, on the 23rd). This will be on the OpenJDK and IcedTea projects which you can find in a GNU/Linux distribution near you either now or in the very near future. We unfortunately couldn’t find any OpenJDK developers in the UK to talk about this, and failed to woo our new OpenJDK ambassador (and I chose that link just because of the way they hashed up his title). Instead, Andrew Haley will come up from RedHat. RedHat have been working with Sun on the OpenJDK project for the last year, and are responsible for the biggest chunk of the IcedTea project. Andrew is the Technical Lead for Open Source Java, though of course we all know and love it as Java Libre or Free Java.

As usual, you can find videos of the demos on my ambassador web site. If anyone would like to mirror these, please get in touch. I’m also currently trying to get these available via BitTorrent too, so any help there would also be appreciated.

It is with a tear in our eye that we said goodbye to the SunRay kit this week. Unfortunately, our plans for promoting this in the DCS just didn’t work out. Our contacts at Sun just weren’t responsive enough in the short time we had, and the whole episode tarnished Sun’s image with the Computer Science support staff.

On a lighter note, plans for our next demo are coming along nicely. Andrew Haley, RedHat’s Technical Lead for Open Source Java, has agreed to come up north in May and speak about OpenJDK and IcedTea, which should be great. This makes a programme of three talks in four weeks: we have an Xvm talk next Thursday (24th April) at 3pm, followed by a Sun SPOT talk on the 30th and then the Free Java talk on the 14th of May. I think that’s about all we’ll be able to fit in, but I think we’ve done quite well in the end. In addition, Sun are trying to organise a road show here at Sheffield. If that comes off in time, I’ll finish my job in a good light.

On that note, I end my stint as campus ambassador on May 31st. I have a successor in the wings, so we’ll see how things go there…

I attended an interesting talk by James Vasile of the Software Freedom Law Center earlier this week. The discussion focused on how the Free Software movement can now be regarded as a success and how others are now attempting to replicate this in other areas such as media (Creative Commons). This is despite flaws in the licenses, and a vast majority not understanding what licenses, such as the GPL, actually mean in practice.

One interesting anecdote James supplied was of a client being confused as to why his BSD-licensed software could be taken and used in a GPL product, from which he could then not used the derived code. He believed that he had a moral right to use the derived code. Yet, the license he chose doesn’t back this up — in fact, this is exactly the opposite intent of the BSD license. The license is incredibly short (only 3 clauses in the newer version) and does little more than retain attribution (via copyright) for the developers. You’ll find BSD code in all sorts of proprietary products, including Microsoft Windows and Mac OS X.

So why does it work? Values like 4 kinds of Freedom defined by the FSF and community are the reason. I’ve always believed, while working with FOSS, that community is central and it was good to see this view being supported in James’ talk. It’s something I’ve also blogged about with respect to OpenJDK. It’s not about licensing or cost, but about being able to give your neighbour your code, and let them play with it, modify it, and pass it on to their friends.

During the talk (and here comes the relevance to my CA blog) I was thinking about how we need to get these points across within the University and that a talk on Freedom and Free Java would be a great way of doing this. With respect to other ambassadors, I believe I’m in a fairly unique position to do this, especially given my recent success as a finalist in the OpenJDK Innovation Awards. At this point, I should thank Tom Marble (former OpenJDK ambassador), who acted on my worries about the usual blanket ‘no Sun employees’ ruling on this contest and got an exemption for ambassadors to participate. I’ve just tested this exemption in signing the eligibility affidavit and queried with Richard Sands about whether doing such a talk would be an issue. The good news is that it isn’t and he thinks it would be a fantastic thing to do. So looks like it will go ahead.

On that note, I’ll close by asking if anyone in the OpenJDK community would also be interested in participating in such a talk/demo session here at Sheffield, and if so, to get in contact with me ASAP.

I didn’t really go into much detail as to the content of the demo in my first (quick) post, so I’ll attempt to do that here. For those who want to see the demo in all its glory, the recordings can now be found on the Demos section of my ambassador web page, along with details of our upcoming demos.

Despite the usual British weather, our first tech. demo went well. We had a good attendance from more than thirty students and staff, which eased our worries about whether our message was getting through. Paul Humphreys opened the session with a short introduction to Sun entitled ‘A Hitchhiker’s Guide to Sun’. Paul showcased many of Sun’s products, including the ubiquitous Java platform, OpenSolaris and an array of Sun hardware including the SunRay technology (which we will feature in a future demo), the open SPARC chip and the ‘Thumper’ (a 48 terabyte storage solution). He also brought along a Sun Fire server containing one of the new Niagara 2 chips, which he opened up and showed to the assembled audience. There were quite a few comments about the lower heat output of the SPARC chip as compared with the consumer x86 and x86_64 chips, which need much more space for heatsinks and fans to keep the chip cool.

Paul then introduced the audience to Jarod Nash, who gave the main presentation of the day on ZFS. He explained how this integrates volume management into the filesystem, allowing individual mountpoints to have a dynamic size allocation. This solves the common problem of a mountpoint, such as /usr, running out of space, when there is still space available on /. With the usual situation where the filesystem is unaware of the volume management underneath, the data would have to be backed up and the filesystems resized. With ZFS, the allocation of space from the pool can simply be shifted. This also allows quota management to be integrated into the system.

The central feature of ZFS is its use of Copy-On-Write (COW) and checksums. In the past, checksumming every operation was thought to be too expensive, but these days the costs is negligible with the speed of current machines. The ZFS approach is to check every operation and ensure that the checksums remain consistent, which basically comes down to not trusting the hardware at all. Jarod explained that while other filesystems can catch bad blocks, there are all sorts of other issues that are missed, usually because something occurs silently in the past and then only shows up later (possibly much later) when the data is corrupt. This can include the disk simply throwing away data, an error in the transmission between main memory and the disk causing the data that is written to be corrupt, or the head of the disk simply landing in the wrong place and writing over some other data. ZFS can catch these errors, and, with mirroring, repair them. Jarod explained that mirroring is essential in avoiding data corruption, as otherwise ZFS has no way to obtain good data to repair these errors. ZFS takes a different approach to mirroring, due to its integration with the filesystem. If an error occurs on one of the disks, ZFS can catch it early, get the good data from the mirror, and then write the correct data back to the erroneous disk as well as sending it upwards to the kernel.

The use of COW means that there is no need for fsck with ZFS. While other filesystems attempt to simply recovery by making an atomic commit to a journal which can then be replayed by fsck, ZFS simply does all writes atomically, ensuring that the filesystem is never in a bad state that fsck would need to recover from. This is achieved by creating a copy of the data, altering that data, and then only pointing to that data when the new version is in place. Thus, if something fails while the copy is being made and written to, the filesystem will still point to the original data, avoiding a corrupted intermediate state.

Jarod closed the session with a live demo of ZFS. Using two USB disks, with a 100mb partition on each, he demonstrated how they could be assigned to a ZFS pool. He then copied data to the pool (the contents of /usr/include) and then filled one of the devices with zeros using dd. By invoking zfs scrub, he showed how ZFS found the bad disk and degraded it. He recovered from this by removing the bad disk from the pool, and then readding it. When readded, ZFS simply assumed it was empty and filled it with good data.

On leaving the demo, students were invited to take OpenSolaris CDs and books. The CDs turned out to be so popular that we ran out, despite Jarod having brought several copies of Developer Preview 2 in addition to our existing OpenSolaris starter packs. Further CDs are on order. The main question that came up post-session was on finding appropriate hardware to run OpenSolaris. Driver support is still not all it could be (Intel and NForce chipsets tend to be best on x86/x86_64), and students were recommended to consult the Hardware Compatability List (HCL) and the OpenSolaris forums.

Today was our first Sheffield Sun tech. demo. In the end, it went rather well I felt. We had over thirty students turn up and a great talk from Paul and Jarod Nash. It’s always worrying when you’re announcing seminars; you never know how many people are actually interested and will turn up. I’m really happy with how things turned out. For those who missed out, I’ll soon be posting the recording of the demo online.

At the end of April, we have two more demos planned; one on Xvm with Gary Pennington and one on Sun SPOT with Bernard Horen. Hopefully, these will be as if not more successful. The Sun SPOT demo should certainly be able to utilise the COM162 students and Gordon’s other contacts to get a good turnout. I need to book lecture theatres for these as soon as possible. This brings me on to my main complaint at the moment, which has been the University’s (and specifically CICS’) lack of support for the campus ambassador scheme. Because I’m only a research student, I need permission to basically scratch my nose as far as CICS is concerned. This is a real pain, as I need to use their aliases to contact an audience on a wider scope than just the DCS. Fortunately, I can get round this with the help of our wonderful admin team, but this doesn’t help any future campus ambassadors. There should really be some infrastructure in place to support the role. I was met with similar issues when I asked for a mailing list. Again, I can get round this with our own great support team, but that’s not an ideal solution.

Acquiring lecture theatres has similar issues. This was aggravated this time by the booking Emily placed on my behalf not going through. I only got definite confirmation we had the room this morning. Hopefully, these are all things that will go better next time. On the positive side, our admin team (as I’ve already mentioned) have been very supportive, especially Monika and Emily, as have our VT postgraduate students, specifically Simon Foster who remembered to organise drinks for the seminar. With all the other hassles, I’d completely forgotten about this, so my deepest thanks to him for picking this up. He also pretty much saved the day by being able to get hold of Paul’s slides from the web via a dodgy wireless connection in St George’s!

If anyone from Sun is reading this and would like to offer to give a tech. demo then you are more than welcome. Please contact me (Andrewj.Hughes sun com) and we can arrange something. I’m especially interested in anything related to OpenJDK, as I feel this has been underrepresented so far in the campus ambassador scheme (which is especially bad given my involvement there). Likewise, any goodies we cdan give out are always welcome (Solaris CDs and books went down a storm this time; we have no CDs left, including all the Indiana DP2 ones Jarod brought). Post-seminar, I was chatting with Simon and Emmanuel Ogunshile in the kitchen, and we thought that it would beneficial to also get a marketing/careers talk in May/June to discuss the possibility of working for Sun. This is also something I need to follow up, and should contact Peter about ASAP.

It’s been a while since I last blogged here, and there are quite a few things to comment on. Firstly, thanks to those who commented on my last blog. The network driver I was attempting to use was for a Via Rhine card and I downloaded it from the OpenSolaris driver community. Both the USB key and the floppy I tried were either ext2 or minix formatted (the floppy was a boot floppy, that’s about the only thing I use them for these days), and the lack of support for these filesystems in OpenSolaris is disappointing. I assume from that that reiserfs is also not supported, which pretty much kills interoperability with GNU/Linux until this is fixed. FAT support is something I’d use rarely. At the very least, Nevada should have picked up the CD but it didn’t and I had to resort to using isoinfo to retrieve the file.

Anyway, since then I’ve had some success. But this was really by giving in. Originally, I really wanted to install a version of OpenSolaris that would be close to the standard Solaris experience of 9 and 10 users (i.e. problems compiling stuff due to the weird compilers, broken shell, etc.) as this would have been a good place to fix GNU Classpath and other Free Java build issues. Unfortunately, the lack of networking with either Nevada or Indiana (which I tried again, but couldn’t even get root access on, presumably due to the passwd bug) I’ve had to use Nexenta (GNU/OpenSolaris) instead. This is much more usable and picked up the network card straight away. However, having a GNU userspace means it won’t really reflect the Solaris userspace that causes so many problems building stuff. Oh well, I’ll just have to stick with our Solaris 9 installs for that. At least the latest release of GNU Classpath I shipped just before FOSDEM should no longer have the readdir_r issue, fingers crossed. The machine is now nicely headless and so I’ll get round to trying things on it soon.

Apart from these woes, I’ve also been preparing for the OpenSolaris/ZFS demo we will host on Friday. I’ve placed appropriate posters (kindly provided by Paul) and sent out plenty of e-mails so hopefully there will be a good turn out. The SAI stuff also came through at last and the details are part of the website I’ve now set up for the ambassador scheme. No idea how much take up there’s been yet though. The delay has annoyed me, as I wasted a lot of time I could be trying the courses myself. I probably won’t get chance to do so until the Easter break now.

On the 13th of February, I intended the Refresher’s Fair to find out more about the Free Software Society (FSS) that has been set up by some of our undergraduate students. I’ve been frequenting their forum for a while. I have to say the fair was a disappointment, and just reinforces my opinion that the Union is a waste of space. How can you deliver information to people when they blast loud music throughout the entire event and squash the stall between two others? It was ridiculous. At least a few OpenSolaris CDs were given out, so someone might have given it a try who wouldn’t have otherwise. We tried one on a laptop at the event, but it seems OpenSolaris’ system requirements are somewhat over-burdensome for some of the commodity hardware our students have. The FSS are running an Installfest this Thursday that I may also try and look into.

Given all these comments about OpenSolaris, it seems I’ll also need some convincing this Friday. So far it’s not shown itself in a good light and the choice of licensing wasn’t a good start. Choosing the GPL-incompatible CDDL means that, while they may be FOSS, elements like ZFS and DTrace can’t be incorporated in to Linux. However, they can be taken by Apple and crippled to meet their proprietary agenda. Yes, DTrace is fine… as long as it’s not used anywhere near our iTunes DRM stronghold that is. There is a ZFS FUSE module for Linux so at least it can be used there, but this is inadequate really. Sun should seriously reconsidered relicensing OpenSolaris under the GPL, especially now GPLv3 is a reality. After all, it was good enough for the OpenJDK, which has already been more widely adopted. Note the appearance of IcedTea/OpenJDK in Fedora Core 8 and Ubuntu, and the forthcoming appearance of IcedTea6/OpenJDK6 in Fedora Core 9.

We have two more talks in the works already; Gary Pennington will be giving an Xvm talk in April, and we are in the process of finding someone to talk about Sun SPOT. It would be nice to also have an OpenJDK talk, which seems to be not too well represented in this whole CA thing (it’s all flipping NetBeans rubbish instead). Definitely going to give that some though, especially with the imminent Community Innovation Awards deadline.

Last week wasn’t particularly a very progressive one, Sun-wise. I once again tried to get a Solaris installation going on the spare Genesys machine and fell at every hurdle. Initially, I went back to the Indiana install I did before Christmas and couldn’t even get root access. I recalled that passwd is dodgy on there; it seems to miss out the letter ‘t’ from whatever password you type.

A Solaris Developer Edition CD (9/07) was handy so I installed that. After being exposed to a gluttony of marketing spiel as it slowly installed (took an age), I got a desktop which wasn’t the OpenSolaris I was expecting. Far from being open, there were globs of proprietary software all over the place (StarOffice instead of OpenOffice, Sun Studio, Sun JDK 6, RealPlayer of all things). It was using UFS as the file system rather than ZFS which I’ve heard so much about. It also still didn’t pick up the NIC.

Trying to get the driver from the web to it was then an interesting issue. How do you get a file to a machine without a network connection? First shot was a USB memory stick. dmesg showed it doing something with the USB stack but no pretty popup GNOME windows or any device nodes (I’d expect popups given this is supposedly the packaged shiny testing version). Next down the line was the good old floppy disc. Again, similar problems. Looking through the kernel directories, I could see very little filesystem support; there certainly wasn’t evidence of any FAT or ext2 support, which you’d expect as a minimum.

Finally, I burnt it to a CD. Now how to read it… Again it didn’t automount it and mount didn’t seem to work either. I ended up extracting it using the ISO tools. Reading around afterwards, it seems Solaris uses some other filesystem support rather than ISO9660, but this doesn’t explain it not working this out for itself. Eventually managing to break open the tarball, the ./adddrv script just stalled (and I left it overnight just to see). Well, I guess we try Nevada, the community edition next. It’s a huge download (~3.6GB) and for some reason they split the DVD into three and then zip each piece, which means I currently can’t burn the image because I don’t have enough space left to combine the three images (although I suppose I could cat them directly into a pipe to growisofs).

Ironically enough, a new Developer Edition release appeared this week (01/08). Even if it is based on a later version of Nevada, that can be achieved separately without all this proprietary cruft. The other thing that’s bugging me at the moment is the whole Sun Academic Initiative (SAI) thing. I know we’re signed up and I want to get started with this competition thing that they were so keen on on the conference call a couple of weeks back, but still no news of the wretched login name and ID! Let’s hope this week goes better.

Good news. It appears the SunRays aren’t just going to be used in kiosk mode. There’s a bigger plan underfoot, it’s just been somewhat delayed. I had a meeting with David Speake, assistant director of technical services at CICS on Tuesday who confirmed that there are in fact 100 SunRays in the IC but also that they will eventually be used for something more interesting.

The plan is to move from a desktop-centric approach (where you login to Windows and use applications) to an application-centric one (where you use the same suite of applications from anywhere). If all goes to plan, the open access centres will eventually have their fat clients replaced by SunRays, providing thin client access to a Windows desktop. But, more than this, the same applications will also be provided via the same technology through a web browser running on any Java-supported platform (in theory — in reality, you’re probably talking Solaris, Windows, Mac OS X and GNU/Linux — at least for now). David showed me how he could load up Excel in this way. This is currently via Sun’s portal software, but eventually it will be part of MuSE.

The reason it’s not been deployed yet is there are still a lot of teething issues and things are taking a lot longer than expected. This is mainly due through CICS having to deploy a Windows server, something they’ve not done before, and link it in to all their existing Novell backend stuff, which involves things like getting Active Directory talking to Novell’s directory services. Fun. And then to add even more into the mix, this all has to interact with the SunRay software on Solaris and be streamed out. At the moment, it works but there are issues with connecting up filestores and annoying dialogs that would confuse users. Hopefully it will all work out.

I also had a chat with Dave Abbott, now head of DCS technical support today about a potential Sun Ray demo. While he’s reluctant to commit to anything (given they already have the fat client approach working relatively smoothly and considerable investment into that with only a small support team), he’s open to having a look at the technology and seems happy for us to use some lab space to setup a few Sunrays and a server. Obviously, any move to Sunrays, should it happen, is not likely to take place anytime soon in the DCS, but I think the use of them by CICS adds significant sway to it. Now I just have to see if the Sun side of things come through with the call on Friday, but things are looking hopeful. I also need to make sure I can find out how these things work myself…

There have been some positive developments recently, if somewhat intermingled with some lows and a little frustration. Last Thursday, things weren’t looking too good. I had a conference call that Peter had organised with his other ambassadors, but couldn’t get through on the number. Instead, Kim phoned me back afterwards and the three of us had a chat instead. As always, I was encouraged by the liberal amount of support I’m getting from Sun (Kim and Pete offered me on top of the already excellent support I’m getting from Paul) but was rather worried about the little feedback I’ve had from the University. Things aren’t helped by the undergraduate students all disappearing about as soon as things really got sorted admin-wise… there’s no teaching until February 11th, and the academics seem to also be using this as an excuse to take things easy for a bit. Pete also mentioned getting similar slow responses, so I’m not alone in this.

After this, Friday was a pleasant turnaround. Gordon suddenly appeared, having apparently read my e-mails about demonstrating and a demo on Sun SPOTs. We had a brief chat, where he demoed the Sun SPOT simulator to me and explained how he’s going to be using it in his module this year. With my Sun SPOT kit having arrived late the day before, we now have three within the department. He said that there’ll also be plenty of interest in a Sun SPOT demo and the biggest issue would be finding a suitable venue to fit all the attendees! Great news!

I also managed to get the phone sorted out for the next conference call on Monday. Turns out that to use an 0808 freephone number the phone has to allow national calls. Tony Chilton actually managed to get this sorted out much quicker than I thought, and it was working by Friday afternoon. I also finally managed to login to the SunRay (by connecting it to the CICS network) and access the remote OpenSolaris desktop within Sun’s internal network. Now I just need to find time to try out some of the resources that gives me access to…

This week, I was introduced further to the Sun Academic Initiative (SAI) via a conference call with Kim, the other campus ambassadors and Nick Beard and David Alderton from Sun. I’m still getting used to these conference calls; having so many people on one call is very weird when you can’t see them or their facial expressions. It seems they are launching a scheme in the UK via the campus ambassadors to encourage students to take the online courses available via the SAI. The CAs can win a prize for getting a high number of courses taken by the students (as well as having the additional advantage that we get free certification).

After the call, I tried this out straight away in the VT lab next door by getting Simon and Henry to have a go at the pre-assessments. Still haven’t been able to get the ID we need to get access to the courses and practice exams themselves though, but I was surprised to see that Sheffield are one of only seven already registered for the SAI. No-one told us, but then what’s new…? Speaking of which, I’ve been fighting with CICS again. It took about five e-mails to get them to admit they were bandwidth-limiting the connection that our research server and the Sunray are on. Then, after I had the VLAN changed, they refused to give me permission to re-register the Sunray so I had to get Mick from our support team to do it.

On the same subject, I went on a mission with Simon back to 1984 (aka the Information Commons) to locate the 150 Sun Rays we are supposed to have. In total, we found 47, all being used not as roaming desktops but as locked down access to two web sites (MUSE and Star) via Mozilla Firefox. They still are the usual Sun Rays of course. If you send yourself a file to your Webmail account that Firefox won’t understand (most things really, but a LaTeX file is pretty safe), then it will prompt you as to a program to use to open it. Opening it with /usr/bin/gnome-session will bring back the full GNOME desktop and a true SunRay, from where you can find that it’s running on a Sun Fire 200 somewhere. Instead of using SunRays like this, they seem to have instead invested in lots of horrible fat Viglen clients running Windows XP. That probably explains why it’s so hot in there in the middle of winter.

Anyway, next week I have a meeting with Dave Speake from CICS so I should find out what their motivation for buying these Sun Rays was. At the moment, it seems to have been just a cheap way of doing what they already do in St George’s Library to provide library catalogue access. I’ve also got another call with Paul and one on Friday about whether we can get a Sun Ray demo setup here. So should be more interesting news next week.