[UKUUG Logo] Copyright © 1995-2004 UKUUG Ltd


Previous Next


Editor's Column

(Susan Small)


Once again the format of the newsletter has changed slightly. At their last meeting, your Council decided to stop sticking the cover CD on the front cover, so you should find your CD elsewhere in the envelope of goodies. If not, contact Jane at Owles Hall and complain!

You should also find in the envelope a "Powered by Linux" badge. These were commissioned by Council and are suitable for fixing to PCs, laptops, etc. and reflect the growing interest in Linux. Once again, contact Jane if yours is missing, or you would like a few more.

Support for the Open Source movement continues to grow and this is reflected in a number of articles. We are particularly fortunate to include a piece from Andrew Hume (President of our sister organisation on the other side of the pond, Usenix) stating why he is "coming home" to UNIX.

According to Council, the June cover CD will be a Linux distribution and the September cover CD will be targeted at E-commerce. If you want something on a cover CD, then let your Council, or Jane, have the details.

Report from the Chair

(Lindsay Marshall)


Well, it's 16.30 on press day for the newsletter and here I am writing the Chairman's Report. My excuse is that I got engrossed in reading the Linux kernel today so I didn't get round to it till now. (I don't recommend reading the Linux kernel BTW - the style is horrendous!)

So, what to say? Well, first I suppose I should castigate those of you who didn't come to our Winter conference. This is a traditional UKKUG event and is really intended as an excuse for members to get together and talk as well as to listen to some technical presentations. Those of us who were there did indeed talk and listen to some excellent presentations. Basically, you missed out so make sure that you are there this year.

Second I should welcome all our new members and encourage them to get involved - the more active members we have the better it is for everybody. It would be good if we could get some more Local User Groups going, but that needs the Local Users to Group. So get grouping.

Hmm, next I should mentiuon our astoundingly successful evening at the Commonwealth Institute with Eric Raymond. We had over 200 people attending (lucky we didn't book the smaller room!) and everyone enjoyed the presentation - even those who knew all about it anyway. The only slight problem was finding somewhere for 60 or so people to eat together afterwards!

We are hoping to repeat the same success with our Richard Stallman talk at the same venue on March 23rd. I hope that you can come along for what promises to be yet another interesting evening, and note that you can get a reduce entry fee to the UNICOM Open Source event that is being held that same day.

What else? Nothing really except to tell you make sure to keep an eye on all the developments in Open Source that are happening at the moment - and there are a lot of them. Oh, and enjoy the newsletter which is its usual eclectic mix of reviews and other stuff.

See you on March 23rd!

Windows Refund Day

(Report by Martin Houston)


February 15th was famous this year for more than being the day after Valentine's day.

It was the day when a number of co-ordinated protests were made across the world.

The remarkable thing about Refund day was not the number of people involved, which was only a matter of hundreds of active participants but the huge media interest that was generated. I myself was interviewed for a TV programme to be shown on Channel 5 which covered the issues raised by the Refund day as part of a wider picture of the interplay between the Internet and Open Source software, Microsoft and issues of Software piracy. As I was writing this the transmission date was not set but please visit my website http://www.deluxe-tech.co.uk where I should have screening information by the time you read this.

What was refund day all about? It was not 'Microsoft Bashing' or an encouragement to piracy.

Simply Microsoft's End User Licence Agreement says (or rather said) in the small print that they can contact their hardware supplier about "returns of unused products for a refund".

With much tenacity several intrepid users managed to get refunds out of their hardware suppliers and, as an exercise in building of the successes of the few, the whole concept of the Refund Day took shape.

You can read about the most influential of these, Geoffrey Bennett's eventually successful struggle with Toshiba at http://www.netcraft.com.au/geoffrey/toshiba.html.

If you want to buy a PC from a major manufacturer and particularly if you want to buy a portable PC, you are forced into paying part of that cost for a bundled Microsoft Operating System. Major hardware vendors get large and undisclosed discounts of the exorbitant 'street price' of Windows OS and Office software. However the EULA says that if the user is unwilling to accept Microsoft's licencing terms they are entitled to return unused products for a refund.

There are a growing number of people now who want PCs but never intend to run a Microsoft Operating System on them. Is it fair that they have to pay a proportion of the price of such a machine as effectively a 'Microsoft Tax'? Although what the computer manufacturer pays for software is not as much as retail price it is still a seizable amount - especially now that 'budget' PCs are going well under the £500 mark. There are two things that people can do about this. One way is to buy from smaller manufacturers. They do not enjoy the attractive discounts that the big boys do and are generally now willing to make Windows a separate and optional item on the price list. Just get a copy of Micro Mart and you will find plenty of manufacturers willing to sell 'built to order' computer systems without Windows.

All very well if you are happy to by "generic" systems. Personally I have built my own systems from parts for several years now. I would recommend it as you get the freedom to choose exactly what you want. However it is not so attractive if you are specifying 100 or 1000 systems. The confidence of going with a well-known manufacturer then counts for more than flexibility.

The problem is that major manufacturers and particularly the manufacturers of portable computers are not yet as flexible as the many thousands of small "generic" PC builders. Even here though things are just starting to change with both Dell and HP now offering systems with Linux instead of Windows pre-installed. If you have been forced to buy Windows because you wanted to buy a specific computer and have never used it you are "in theory" entitled to a refund.

"In theory" as to actually start giving refunds on a large scale would land big troubles for Microsoft.

How much should the refund be? Revealing that would be revealing the highly secret discount structure.

Refund day is more a point of principal than a serious attempt to get £50 or so refunds for the 10 million plus Linux users. Also there is the situation where systems have been sold with bundled software "worth hundreds of pounds!" both from Microsoft and other vendors but all useless without Windows. How much is such a bundle really worth? For several years now many manufacturers have depended on big software bundles to give themselves competitive edge. How many of you have bought PCs with bundled software little of which actually saw any use?

By the way I am not myself seeking a refund. I do use my copy of Windows but only about 10% of the time. The more software gets ported to Linux the less I will have to spend in Windows. 10% use did not justify upgrading to Win98 nor will it justify going to Win2000.

Microsoft claim that the refund day was just a publicity stunt. As a stunt it was highly effective with a high level of news coverage both before and after the event. The media are now ready to listen to the message that there is more to the future of the computer industry than successive versions of Windows and Office.

The thing that is most damaging to Microsoft out of all this that as a company that is so hot on holding others to the terms of contracts and licenses should be caught out with a term in an agreement that hundreds of millions of people have supposedly agreed to, that Microsoft itself has no intention of keeping. Is the Windows EULA really worth the paper it is printed on?

The best thing that Refund Day can achieve would be to allow all PC buyers to opt out from being charged for Windows if they do not wish to use it. If manufacturers want to temporarily licence Windows to test that a system works with it, then it is up to them to come to a suitable agreement with Microsoft for such a use. Blindly assuming that the buyer is willing to accept and carry on with the licence has to end.

You can find out more about the Refund Day at http://www.linuxmall.com/refund

News from Owles Hall

(Jane Morrison)

Jane We are very pleased to announce that the 1999 subscription payments have been coming in at a steady rate, thank you to all our prompt paying members!

For those who haven't paid yet we shall be sending out the first reminders next month.

The UKUUG Council first meeting of 1999 was held in London on 9th February. Discussions on future events for UKUUG were top on the Agenda and a full programme is envisaged for 1999 and 2000.

You should find (fingers crossed) a Powered by LINUX badge in with this Newsletter mailing - we are sending a badge to each UKUUG member - please let us know what you think about the badges and if you require any more!

The CD this month is also enclosed in the envelope instead of being stuck on the front of the Newsletter - it was thought by Council that you would prefer this method which means you don't have to rip the front page to get the CD off!

We are pleased to announce that the Eric Raymond evening talk held on 20th January was a great success with more than 200 attendees.

The next UKUUG evening meeting will be held on Tuesday 23rd March at The Commonwealth Institute, Kensington, London, where Dr. Richard Stallman will give a presentation from 7.00p.m. - 9.00 p.m. entitled The GNU Project and the GNU/LINUX System - details have already been circulated to our members - please note that there is no need to pre-book and a booking fee is NOT necessary.

UKUUG Winter Conference 1998

Using those Spare Cycles

(Report by David Hallowell)

David The UKUUG Winter Conference had been held in the North for the past few years so it was decided to bring it back down South for the 1998 event. This was the first winter conference I attended and it meant a lot of travelling for me from Newcastle. However, the one o'clock start meant that I had plenty of time to reach Essex.

The Wivenhoe House Hotel, where the conference was held, is located in the grounds of the University of Essex. It is an 18th Century manor house surrounded by extensive parklands; it was only a short bus or taxi ride away from Colchester's main railway station.

The conference was spread over two days with a conference dinner and lunch available. This is keeping with the tradition of previous winter conferences (so I believe). The conference was organised by former UKUUG chairman Mick Farmer.

After registration and a chance to meet the other delegates the conference got underway with Savas Parastatidis from the University of Newcastle upon Tyne with a paper entitled Implicit Parallel Computing on Networks of Workstations Using NIP. The paper describes NIP which Savas and Paul Watson are developing at Newcastle University.

Parallel architectures based on networks of commodity hardware workstations are proving to be cost-effective platforms for the execution of parallel scientific applications. Systems like Avalon have demonstarted that it is possible to have vast computational power only for a fraction of the cost of a customised supercomputer. However, if networks of workstations are to become widely used as parallel systems, programming methodologies that ease the development of parallel applications must be used. Many believe that the commonly used message passing methodology is too complex and time consuming and puts too many responsibilities on the programmer. Programmers have the responsibility for making the important decisions on how to exploit parallelism. This has major advantages and requires specialist expertise to become a parallel programmer.

For this reason the design of an implicit parallel system has been investigated in which it is the compiler and run-time system, rather than the programmer, that are responsible for the creation and management of parallelism. The run-time system, called? NI, has been designed and a prototype of it has been implemented. NIP removes from the programmer the burdens of dividing a program into parallel tasks, sharing work evenly across all nodes, organising communication between tasks, making the data available to tasks that require it, and synchronising tasks.

A very detailed paper discusses parallel programming languages, the parallel machine architectures, an overview of NIP design, the NIP execution Model and the Distributed Shared Memory Model. The details of which are too extensive to reproduce here.

There are plans to extend NIP in a number of directions. First, the behaviour of a wider range of parallel programs will be analysed. A set of advanced caching features are also being designed to improve the efficiency of the NIPDSM (NIP Distributed Shared Memory system) by copying and caching groups of objects rather than single objects. Finally, they're investigating the mapping of other high level languages onto NIP; the initial focus is on extracting implicit parallelism from Java programs.

After a short coffee break Peter Lister presented his paper entitled Persuading Professors to Purchase a Pile of PCs which outlined his experiences of constructing a Beowulf at Cranfield University. The idea to construct a Beowulf arose as the university needed a powerful system with a limited budget (£150K). A Beowulf meant they could construct a powerful system with cheap Intel hardware and Linux which is free. Then only needed 20% of their budget to construct the Beowulf which meant that the other 80% could be spent on other hardware.

The main problem was some of the professors were doubtful about the success of a Beowulf and needed persuading (hence the title). As only a small amount of the budget was going to be spent on the Beowulf it meant that the professors that didn't think that constructing a Beowulf would be a good idea as the money good be spent on more proven hardware (such as Alphas, Sparcs, etc) could be reassured that there was still 80% of the budget available for them. For those that thought a Beowulf would never work they were reassured by the fact that as it was ordinary PC hardware they could easily trade the hardware in or use the machines as ordinary desktop PCs or Linux file servers. For those who were in support of the Beowulf but thought it should be bigger were told if it was successful they could apply for funding for a bigger system.

Once the professors were persuaded the task was to buy the hardware and set up the Beowulf. Peter's motto was Don't be proud - scrounge and he contacted dan, their supplier for Intel based hardware, who offered him the option to return it after 3 months if it all went wrong or upgrade at low cost if it was a success. The set-up consisted of a master machine and 16 slaves (all 300MHz Pentium II's with 192MB DRAM). The operating system used was Red Hat Linux 5.0 (Red Hat now have an Extreme Linux distribution designed for Beowulf's but this setup pre-dated this) with the slaves configured with Kickstart. Other software was LAM 6.1, PBS (for batch queuing) and a mixture of applications developed in-house.

They did suffer problems with the hardware. The PC's used AMIBIOS which insists on having a keyboard and video card for the machine to boot up (if you plan to construct a Beowulf make sure that the BIOS in the machines you plan to use don't require a keyboard to be connected unless you have plenty of spare keyboards) and when the memory in the machines was upgraded a lot of errors started to occur - they insisted that all the memory in the machines to be identical and the problem went away.

They had a few problems with Fortran compilers (try before you buy) and a few problems with LAM but in the whole the problems were minimal.

The benchmarks for their Beowulf (they called it The Borg) were impressive. The benchmarks compared favourably with an 8 processor SGI Origin which would have cost £120K compared to £30K for the Beowulf.

The third speaker was Laurent DeGrouve who is a representative from distributed.net. On the distributed.net web page they claim to be The Fastest Computer on Earth. distributed.net consists of a network of thousands of volunteers around the world who run a special client on their machines. This client utilises the idle CPU time of the machine to tackle the current project. At the moment distributed.net are working on cracking a message encrypted with an RC5-64 key, previous projects have involved cracking DES and RC5-54. Laurent described how distributed.net works both at the client level and at the server level which involves receiving completed keys from the clients and sending new keys for the clients to crack. An important part of distributed.net is their stats server where you can check the overall progress in the latest project or your own individual progress and if you've joined a team your teams progress.

The current and past projects all involved cracking some form of encryption but future projects are designed to be more varied. The next project planned involves finding Optimal Golomb Rulers which is the term given to a set of whole numbers where no two pairs of numbers have the same difference. For more information on this future challenge see http://www.distributed.net/ogr/

Possible projects for the future include a distributed chess engine, finding large prime numbers, scanning radio signals for extra-terrestrial life, finding fermat numbers and probably more cryptography projects.

Anyone is free to join distributed.net and become a part of the fastest computer on Earth.

A 30 minute panel session followed which gave the delegates a further chance to ask the speakers questions. This panel session was the end of the first day's conference proceedings. Most of us made our way to the Hotel bar which gave us another chance to meet the speakers and other delegates. This was followed by an excellent dinner at 8pm. It was an ideal end to the first day.

The conference resumed the next day at 10am with a paper enitiled The Ages of Data by Phil Davies who works for the European Radio Development team at Nortel Networks. The paper addressed the problems they suffered with increasing disk usage. A sizable amount of their budget was spent each year on increasing storage capacity as the demand for disk space was growing exponentially. The problem is that it is difficult to know exactly how to control the problem. A commonly suggested solution of imposing strict disk quotas could mean people delete data to free up space that may turn out to be useful in the future. CAD programs are particularly guilty for taking up space as they weren't designed with efficiency in mind or indeed for multi-user systems. The CAD programs often store their libraries in the users area meaning there's a lot of duplication, space which could have been saved if the CAD programs were better designed. Another idea was to delete projects over two years old which were unused. However only 5% of the disk space was used by projects over two years old which would save so little disk space as to be worthless. However, if the projects were deleted within a few months of completion there would be the opportunity to save between 1/3 and 2/3 of the disk space. This, however, would severely limit the potential of reuse of designs. However if 50% of the data over one year old was deleted and all of the data over two years old there'd be the potential to save 25% of the disk space. Although this was not a significant amount it was certainly worth considering.

Hierarchical storage offers the best of both worlds. It keeps constantly accessed data online and moved older data to cheaper but slower storage. It allows design reuse without the large cost of online disk. However for a small to medium sized data set, the cost of the hardware to impliment this far outweighs the savings.

In an environment where the opportunity to reuse designs is low then archive to tape is a reasonable solution. It is an extension of the backup process and should be easily implemented. However if files are regularly required to be retrieved the time delays and system management time may well cost more than the cost of keeping older disks online.

Their chosen solution was to retain the previous server keeping data available for as long as needed. The disk and tape size would be approximately doubled each year. Gentle quotas would be applied which made the disk always appear 90% full and have occasional but aggressive 'tidy ups'.

The UKUUG vice-chairman, Simon Earthrowl, presented a paper entitled Trend Spotting. The paper presented ideas for the improved management of UNIX type systems, using simple tools that are already available for system administrators. He presented ideas of creating daemons to automatically perform tasks that are normally routine and tedious. A small amount of effort can save a large amount of repetitive work in the future.

The final speaker was Adrien Belcourt with a presentation entitled Internet Access Revenue Flows. As its name suggested the presentation discussed the various ways of making money out of providing Internet access. The presentation compared how the situation differed in the USA compared to the UK and Europe. Of particular interest was how each country's percentage of the population with Internet access varied. This was often down to a number of factors such as the way ISP's charged for service or telephone charges. In some European countries ISP charges were so high it would work out cheaper for light Internet users to dial up a free UK provider such as Freeserve. In these areas there's room for competition.

Unsurprisingly the USA has a very high percentage of the population with Internet access because of the free local calls that just about every telephone company provide. The only source of revenue in the USA is the subscription fees and with the competition of ISP's these generally aren't very high.

In the UK you're extremely lucky if you've got a local Cable company that offers free calls to ISP's (even though most offer free local calls to other phone customers). This means that there are two possible ways in the UK to make money out of Internet access. The first method is via the subscription. Most people in the UK still pay a subscription to their ISP's but once the free services improve more ISP's will start offering their services free and many small ISP's will die out.

Free providers like Freeserve are registered as a telco so they get a share of the call charges to their service. This is expensive to set up but means it's possible to provide a service that appears free to its subscribers but they still can make money. It's expected that more ISP's will follow this route in the future. So the UK seems to be going the opposite way to the US.

Proceedings were running about 10 minutes late so it was straight to lunch. The lunch was a three course buffet style meal with an excellent selection of food. That marked the end of the conference and the end of an interesting two days. The venue was excellent and there was a varied selection of speakers.

Previous Next
Tel: 01763 273 475
Fax: 01763 273 255
Web: Webmaster
Queries: Ask Here
Join UKUUG Today!

UKUUG Secretariat