Windows Vista

[TheMonkey]

TCP/IP was completely rewritten in Vista, in 2k3 it should be about the same as in XP - think minor changes but that's about it. Don't blame MS for having crap TCP/IP - they got it from BSD...

Ironically, look at my previous post. FreeBSD wipes the floor of 2k3.

 

[AtW]

Ironically, look at my previous post. FreeBSD wipes the floor of 2k3.

Yes, but that's because the TCP/IP stack that Microsoft took was at least 10 years old, no wonder current FreeBSD's one wipes the floor. I have high hopes for Vista's TCP/IP, they did not hype it that much so I think the changes are probably true, they are just not sexy to hype.

 

[DaveB]

Yes, but that's because the TCP/IP stack that Microsoft took was at least 10 years old, no wonder current FreeBSD's one wipes the floor. I have high hopes for Vista's TCP/IP, they did not hype it that much so I think the changes are probably true, they are just not sexy to hype.

If you take 10 year old technology and expect it to perform any better than it did before withoutactually doing anything to bring it up to date, then yes, I'd say you are to blame for it.

MS knew they had a crap TCP/IP stack, they just chose not to do anything about it till now.

 

[TheMonkey]

For which they have suffered and now we're going to suffer due to a buggy stack for a few years.

Apparently the old stack memcpy's everything in a packet as it floats down the OSI layers. Piss poor. Everyone else just passes a ponter to the packet.

 

[AtW]

If you take 10 year old technology

It was probably 15-20 years old actually, they took it 10 years ago and back then it was in use for a long time.

Intel is preparing chipsets that will accelerate TCP/IP - this should provide enough kick to actually have available bandwdith being the main bottleneck.

 

[DaveB]

It was probably 15-20 years old actually, they took it 10 years ago and back then it was in use for a long time.

Intel is preparing chipsets that will accelerate TCP/IP - this should provide enough kick to actually have available bandwdith being the main bottleneck.

And so the cycle continues. Crap software, cant be bothered to fix it, just chuck some faster hardware in instead. Faster hardware means we can write crapper software untill the hardware can't cope. So get some faster hardware.

 

[AtW]

Following your logic we should not have hardware accelerators for 3D apps?

Hardware is in reality software, just in hardware implementation - TCP/IP should not require special hardware per se as general purpose CPU should be able to deal with it, but new opcodes specifically designed to help would improve situation - SSE4 by Intel actually looks good, particularly CRC32 opcode - that should save plenty of ticks for TCP/IP.

 

[Joe Black]

And so the cycle continues. Crap software, cant be bothered to fix it, just chuck some faster hardware in instead. Faster hardware means we can write crapper software untill the hardware can't cope. So get some faster hardware.

Hey, that's my way of thinking.

Which is cheaper, paying an IT consultant squillions per day to optimise the software, or (after Intel has spent a few bil on R&D) just reap the benefits and purchase another 1GB for fifty quid?