With WAN optimisation vendors such as Expand, Riverbed, Juniper, Blue Coat and Packeteer developing PC client software, I reckon it's time to create a standard for the PC end.

If the average laptop is now powerful enough - and the software evolved enough - to do its own protocol optimisation, perhaps some of those optimisation capabilities should be built into the system stack.

After all, in the early days of the PC, users had to choose, install and maintain their own TCP/IP stacks. Software companies such as Hummingbird, FTP Software and Trumpet made a nice business out of adding TCP networking to Windows 3.0 and 3.1 - until Microsoft recognised that it was now a must-have, not an optional extra, and built its own TCP stack into Windows 95.

Microsoft's move destroyed those third-party business streams in one fell swoop. Could the same thing now happen to the underpinnings of WAN optimisation? After all, most of us now run remote applications of one sort or another, and problems such as WAN latency, TCP inefficiency and slow file transfers are pretty much common to all.

The WAN optimisation companies are resisting, of course. They cite the value of their intellectual property, and how their application optimisation routines are better than anyone else's.

But a standard way to deal with latency and TCP inefficiency would surely help everyone, not just Windows users. And just like having a standard TCP stack, it would stop the WAN guys continually reinventing the wheel and free them up to work on areas where they can really differentiate themselves.

The challenge for the vendors have to face up to is that if they don't do it - or at least sponsor a group within the IETF to do it - they risk being abruptly shut out by Microsoft.

Even worse, they could be trumped by Microsoft in alliance with one of their hardware rivals, such as Cisco.