How to Really Scare Microsoft
Back in November of 2005 I attended a talk by Marcus Ranum entitled "How to Really Scare Microsoft". The slides from his talk are available online. He talked about a variety of topics which is why my notes from the talk are all over the place. And these notes reflect a fairly unfocused note taker too.
It isn't so much about how to really scare Microsoft, but it should be how to kill system administration?
Startups are all about vision.
Is tenable network security something that can be achieved? Right now we have firewalls, VPNs, and IDS, but there are still problems.
Microsoft is no longer a good custodian of the industry. It is still a young industry, but with many failures: Cray, Digital, Data General, and Wang.
Someone should start a dead pool on Microsoft and Sun.
The reality is that business aren't switching from expensive and mediocre to free and good.
Linux is "trying" to match Windows. Ranum likes regedit more than termcap.
"Avoid strength, attack weakness" -- Art of war
Linux should be attacking integration, new features, software distribution, 3rd party software is bane of system administrators.
IBM backend server market.
How to attack Microsoft?
System administrators have a hard time setting up and configuring a Microsoft environment. SMS, AD, etc.
Microsoft has data lock-in, too many document in Office format.
Change the software sales cycle, something like cell phone model.
System administrators are the achilles heel of general computing.
Appliance computing is a possible killer.
Microsoft can't ease system administrators since they are so ubiquitous.
#1: Microsoft can be attacked by inroads in system administrator costs
Microsoft monopoly trial was a hit by Sun and Oracle
Microsoft products divided up by those that are in, out, and sold separately.
Software sales cycle needs to switch
#2: Freeze IT spending for two years
Can't step off treadmill of code updates
Simplicity is a virtue.
Different sale model.
Manage by using slightly old hardware and software.
Steal ideas from old research.
Co-opt the open source ideology.
Main lines: system administrators, cost per seat. performance, reliability, mobility, security.
1) Data environment, standard XML, and standard applications.
2) OS bloat from virtualization. Make everything a file with properties. PGP everything. Everything exists in multiple places, cache as much as possible, file service is a software distribution model. Multiple copies, multiple versions. Everything has a URI. Keyring in a friends' computer lets you access anything.
3) Platform (PS2/PSP).
OS upgrade by NetFlix.
Same things over and over again but in different ways. Build new things, instead of rebuilding old applications.