On working late:
I'm a fan of the XP approach - that people should work 40 hours a week. I think 40 hours of focused work combined with time for other interests, friends and sleep is more productive than 60 - 70 hours and nothing else. There are occasional instances where we ask for weekend work and the like but it's two or three times a year.
That said, programmers are adults and I let them manage their own time (so long as they're doing their minimum contracted hours). I've politely suggested to people that they really don't need to be working late when I'm getting mails at 11pm and there's no big deadline but generally it's up to them.
On Parkinson's Law:
This is oft quoted but I'm not sure too many people have really looked at it's origins and whether it's really applicable. The phenomenon was primarily observed was the British civil service in 1958 and then in government organisations since then and related to environments where empire building and a lack of urgency were common place.
Personally I'm not sure that that will really tell us too much about the workings of modern IT functions where there tends to be a drive for efficiency and continual pressure and deadlines which combat much of what it describes.
It's probably more true in very large companies to an extent, where the necessary bureaucracy can take hold, but I don't think the motivations it talks about (essentially empire building) are things that programmers are driven by and therefore it's impact at the level where work is actually done should be minimal. Of course projects may become mired at the analysis and management levels where those factors are applicable but that's a different thing.
So I don't really buy it for most IT organisations as it's originally formulated.
I think there are problems which lead to something which looks like Parkinson's law (essentially people are given tasks which are too large, don't apply themselves at the beginning thinking they have plenty of time and then have to work like crazy at the end right up to the last minute unnecessarily), but that largely comes down to poor management (by the PM who should have set interim deadlines) and time management (by the programmer who should have buckled down at the start).
Edit: At Pierre's prompting this summary from what Peopleware says about Parkinson's Law from http://javatroopers.com/Peopleware.html#Chapter_5:
Newton's Laws have stood the test of some 200 years of subsequent study. Parkinson was a humorist, not a scientist. His "law" caught on because it was funny, not because it was axiomatic. His examples were observed in a fictitious British government bureaucracy, where they give little job-derived satisfaction. The simple truth is:
Parkinson's Law almost certainly doesn't apply to your people.
Their lives are too short for loafing on the job, as that would delay the satisfaction they hanker for. They are as eager as managers to get the job done, provided they don't have to compromise their standard of quality.
"You Wouldn't Be Saying This If You'd Ever Met Our Herb" - In a healthy work environment, the reasons that some people don't perform are lack of competence, lack fo confidence, and lack of affiliation with others on the project and the project goals. They are overwhelmed by the diffculty of the work, pressure won't help, reassignment, possibly to another company might. Treating your people as Parkinsonian workers certainly won't work, it can only demean and demotivate them."