What the hell? So when a driver of a Toyota car crashes into a tree it's Toyota at fault as opposed to the driver?
I have no idea why the author has decided to crowbar their own bias on platform preference into this article. There doesn't appear to be any information that states its the platform at fault.
Did Nasdaq get rid of the hp/tandem nonstop backends then? I don't think they did. As of 2008 they were still running them. They might have windows frontends and middle ware, though.
My first job in the Silicon Valley was on the Himalayas doing QA. They were awesome machines. QA meant pulling drives and cpus in the middle of processing, browning out/cutting power, etc.. it was wonderfully destructive :-)
I stopped reading at the point he described the existing solution as "a custom blend of C# and .NET programs." He obviously doesn't have a clue about the platform he's bashing.
I mean, I derive no pleasure from using bloated MS software myself (cough Outlook cough), but this guy clearly has an agenda that guides his reporting.
Agendas can be Ok, if you have a clue, and some idea of where the real story lies coughAccenturecough. Where I would start - What was the spec for the system? Did it change? What development method was used? was it agile? What test coverage did the system have? How many developers were on it, and how experienced were they?
Leaping from "a windows-based system failed" to "anyone ever fool enough to believe that Microsoft software was good enough" and mentioning .Net and MS SQL Server as specific examples is just moronic and ignorant.
TradElect was launched at the same time that BetFair, a big internet betting exchange, announced its own FlyWheel platform. FlyWheel took 2 years to develop internally (apparently with two different dev teams initially competing against each other) at a cost of around £1m. It runs on Oracle/JBoss/Linux/x86 on £25k worth of kit, and does up to 1m TPS (compare with TradElect's £40m to achieve a target of 68k TPS, which apparently it never achieved).
I heard about this. It's a real shame that BetFair's system isn't being implemented for the LSE- i hear it far out strips any 'real world' use cases and can handle it all really well. Sounds pretty awesome to boot. :)
The article doesn't seem to indicate that LSE will be dumping windows -just dumping a third-party product that happened to run on windows. It is unclear what platform the new vendor will be using. It is also unclear that the trouble outlined here was the fault of the OS or just a poorly-coded program running on top of it.
I can't imagine why anyone would ask Accenture to program something as critical and technically demanding as a trading platform. Their public record here in the UK is appalling - the NHS and British Gas are two horrendous cock ups that spring to mind. How on earth do they win these kind of contracts? Brown envelopes?
Absolutely. You know who writes code at Accenture? 22-year-old liberal arts and business majors, in their first job out of college, who've had about 6 weeks of training in programming.
There is a great comment in the article from a person who worked at Accenture...
"I finally quit after walking out of a berating meeting from a boss who reckoned I'd done a project too fast and produced an executable that was too small. "We quoted them three point two million for this. We can't give them a 500 Kb executable after a week and say it's done"
For 3.2 million, I'd hope for some good documentation, guides, and some fanfare, preferably some evidence of the QA procedure it's been through, too. You can't put ten thousand dollars+ of polish on a project in a week, can you?
hmm... I was surprised several years back by the announcement when LSE (London Stock Exchange) moved to MSFT server products, since people in banking mostly tell me that Java/Unix is standard.
There is no "standard". In any investment bank you'll find everything from MUMPS to MATLAB, VAX Pascal, APL, Haskell. The result of a) many acquisitions and mergers and b) a culture of if you need something, build it yourself, out of whatever you have to hand. I've seen NeXTStep and OS/2 and even an Amiga on trading floors.
Wait a sec, that's not infrastructure there running in individual trading departments. Major transactional backbones in banking almost all run on Java/Unix as I understand.
For the really mission-critical systems there's more COBOL, C++ and even VAX Basic (I'm not kidding) than there is Java, and there's a lot of mainframe and VMS too, and weird Unixes like Dynix. A lot of it is stuff that people outside the industry will never even see anymore.
One project I worked on in fixed income was C++ based, it could handle a few hundred trades per second on modest kit. When I left I think the J2EE rewrite was up to 6 trades per minute...
Thanks for the insights! Yes, my friend's dad in Munich used to program the main software handling Hypo-Vereinsbank's transactions and it was all in COBOL (a previous programmer had introduced a small rounding error, which resulted in havoc... good story).
Very interesting to see this diversity of systems.
Transactional backbones run COBOL, not Java. Only the inter-system interfaces uses Java and other "new" languages. However, most of them run on Unix...as in, versions of Unix that are older than the internet.
My painful discovery of the day was the sad revelation that in 2009 Microsoft still has not yet moved on from case-insensitive filesystems. I'm sorry, but if your operating system cannot distinguish between ['ass','Ass','ASS'] then it is a toy and not a tool.
Modding me down does not erase this painful truth.
A case insensitive filesystem needs to convert everything into a single case before doing any string comparisons. The only reason Unix doesn't do this is because they couldn't spare the CPU to do it in the original implementation. So not only are you offtopic, but you have your chronology backwards.
Sorry, but a filesystem that doesn't preserve exactly what I put into it is broken.
And the only chronology I mentioned was my surprise that Windows is still broken, even in it's latest iteration (Vista).
I am glad that I only have to deal with Microsoft product on an extremely occasional basis. I feel sorry for people who are so emotionally invested in their operating system that they can't see that OpenBSD is the answer... ;-)
> Sorry, but a filesystem that doesn't preserve exactly what I put into it is broken.
I work with non-hackers, non-computer people who use computers as tools. They are constantly surprised that Linux does make a distinction between cases. They remember the words they used as the names of documents, letters, reports, etc., and they don't want to have to remember whether it was capitalised or not.
Dictating names of documents becomes a chore when you have to specify the capitalisation. Just the other day my wife was trying to find a submission of mine here on HN and ended up phoning me in frustration, exactly because the system remembers/keeps exactly what was put into it.
Your attitude, quite rightly here for a hacker, is that you want the system to do exactly what you want. For entrepreneurs and others who deal with non-hackers, knowing that most of the world don't see things the same way, and occasionally want something "obviously wrong," is important.
Windows does preserve exactly what you put into it, it will never change the case of your files. It will not let you call two files the same name (with different case) and it will not care if you reference a file with the wrong case.
This is utterly brilliant behaviour. Case is a pain to remember and a pain to communicate with other people, it doesn't fit well with our language. You only need to try and read a password to someone to see that. There's also no pleasant way to remember that two files are the same name but different cases.
One of the first things I do when installing linux is set bash to case insensitive tab completion, and vim to case insensitive searching.
You may as well ask for an OS that preserves the speed you typed a filename and wont let you access it unless you match the speed again.
Actually, this is not entirely true. If you use a program called PAX, which is like unix tar, you can get two files with the same case-independent name with different capitalization. This is necessary for posix.
And yes, this is very odd, and apparently not well known. Just as is the fact that from the C-level api, you can use forward slash instead of backslash to separate file name and directory name.
Perhaps the smiley wink ;-) at the end of that sentence might have clued some people in to the fact that it was not meant to be taken completely seriously.
I am both heartened and disappointed by this thread, heartened that a few people came up with reasoned arguments as to the usability benefits of a case-insensitive OS. Disappointed at the far larger number of people who hit the downmod button on comments they disagree with without being able to articulate why they disagree.
I think that the latter behaviour is all too representative of Hacker News of late, unfortunately.
I have no idea why the author has decided to crowbar their own bias on platform preference into this article. There doesn't appear to be any information that states its the platform at fault.