Seriously? It's gotta be Kongbucks....

Hang on, no groklaw.

Damn

I want remote brick, if I lose my phone, I want it being completely useless to the next person, no firmware flash, no nothing; a paper weight. I don't want it being sold off for a tenner and sent to another country that doesn't subscribe to the block list.... Actually, you know what? I want it catch fire, I want it to be an incinerated paper weight!

> You're absolutely right, they weren't designed right.

Well, they were designed right back then when markets behaved the way they did. Things change. I know people on designing medical software now that wouldn't design it the same way 5 years ago.

Finance is no different. 10 years ago, you would never have considered Hadoop or anything like that, but now, large distributed systems are exactly what people are looking at instead of running batch all the time.

> I thought old, inappropriate systems at banks were common knowledge...

Only to people who work in IT / Finance, and only more recently with the failures in the UK and other places which showed how antiquated systems are.

It's more a management point of allowing what is happening to happen. "Why not push the HFT customers off to another platform so that the rest of the customers aren't impacted by them?" was more my point.

Let me get this straight...

> They can add sometimes unexpected ramp-ups in data that can cause already-creaking systems to fall over

So, your systems suck.

> potentially impacting other non-HFT clients. This is partly bad management of older and less interesting systems but partly because they are an unpredictable lot.

Partly?!?!?!?!?!

Sheesh, if you can't handle ramp ups and ramp downs, your code / systems aren't designed right. Screw the HFT guys, they'll be using other people aside from yourselves to get to market when you stuff goes down / goes slow. What about your other clients? These are apparently your more profitable ones and you're letting the HFT guys break the platform they trade off?

Ha. No. Where do you think they get their traffic data from for Google maps?

RHEL works fine on my media centre with EPEL and nux-desktop. It gives me everything I want and more, and I don't have to worry about the OS falling out of support after a couple of years...

But the only reason to not run a 64bit OS and 32bit apps is because you need to run 16bit code.

PAE won't provide a big system cache because the kernel is still 32bit, it can't address more than 4GB itself. A kernel won't be able to address itself through a PAE window.

> I'm not an expert

No offense dude, but, you're not. This isn't going to address the issue. The problem here is running 16bit apps and mixed apps with 16bit code bundled in with them.

PAE is physical address extensions. This means that you get multiple "windows" of RAM which means that you get to switch your view on which window of RAM address space you get to see. This allows you to see above 4GB which still using 32bit address space.

This doesn't allow 32bit applications to see more than 4GB natively, it doesn't allow the kernel to address more than 4GB (It's still a 32bit app after all), Citrix boxes can't magically take 20 times the amount of sessions, it's for large data storage applications.

So, if you have a 32bit app, it can only natively address 4GB of RAM, with the divide with Windows, this is half the amount of RAM you've got installed minus drivers. Graphics cards which have 1GB of RAM, will take 1GB out of the system so that the graphics drivers can address the RAM on the card.

This is great if you are MS, Oracle or VMWare and are fine with writing your own memory manager but writing a memory manager for allocations is a pain.

The poster was looking for a way to get old apps to work on a new OS, and honestly, getting it to happen reliably is a PITA.

If this is being done for a business, it's a bit of a dark art, all I can suggest is test test test test and test some more.

They could have sold a few million of those things, everyone was raving about it, and then they killed it stone dead. Even though it had a MS badge on it, I was willing to give it a go.

I have a feeling that Steve Balmer is out of touch, or maybe I am, I don't know.

There is a cost of change.

The UK banks are a prime example, there is so many inter-dependant systems, not only within the bank, but also with everyone who they interact with. This all needs to change; in NZ with a smaller population, smaller number of customers and subsequently smaller systems and interactions it's a LOT easier to be more nimble.

Wear.

Contact chips used in chip and pin cards have to be regularly replaced. Contactless cards last... I'd nearly even say indefinitely. I've had my current Oyster card for 8 or so years now (Lost my last one when I lost my wallet) my chip based cards need replacing at least every 2 years.

Something in me thinks that we've been down this path before....

It all comes down to who's watching the watchers....

Linux + SELinux, (SELinux, which was originally built by the NSA for those who don't know enough history to realise) is an operating system with an immutable watchdog. What more do you want?

If you have the source code and the policies, both of which can be externally audited, how can you (As an external person) screw this up?

I remember back in the old old Solaris days dealing with buffer overflows in the driver stack to get remote root, but those days are gone, you would never get that permission to access that executable, let alone open a socket.

If you've got SELinux + policies it's here and it's here now.

Just in case you think this is a pro-Linux rant...

Microsoft have spent a truck load of money on "trustworthy computing" to find new exploits, to the extent that they have honeypots to find new stuff for back testing.

They don't have a watchdog yet, they've started with Windows Defender, but that's nowhere near low level enough yet, and the whole anti-competitive landscape, plus developer buy in (And unfortunately a lot of devs don't know exactly what they're really doing) makes it difficult to say the least. They are still a couple of OS released away from making it work.

Don't discount, and I'd nearly go so far as to say, ever. It makes you look cheap.

The moment you offer a discount is the moment when they start thinking they aren't getting a good enough deal and they'll screw you more or they'll resent it.

The only possible exception is if your contracting and someone is buying hundreds and hundreds of hours of your time, at that point, you don't have to worry about dead time so much (You should have this factored into your hourly / daily rate) and you can take that out.

Where's grub when you need him with a good Dr Bob post?

My coworkers looked at me like I was nuts when reading through that post.....