So why a honeypot and not traffic monitoring?

I don't think the GP is saying that at all.

It would just be nice to have a bit of balance. I don't think we should scrap online anonymity because some of its uses, in opposing tyrannical powers, whistleblowing and similar, are too important.

But we should still recognise that it has costs. Anonymous online abuse can and does have very serious consequences, including depression and suicides. Too often the argument is presented as one-sided.

That's a pretty naive view of what lawyers do and when they're needed. A lawyer is just an expert in certain things, and they assist people who need to do something they aren't themselves an expert at:
- If you want to present a persuasive case in court, you want someone who is an expert in reviewing and presenting evidence. A lawyer can be that person
    You also want someone who can explain the case to the court in a persuasive way. A lawyer can be that person.
- If you want to prepare a contract that does what it's meant to you need someone who is an expert in precise writing. A lawyer can be that person.

Moreover your points about society are just flat-out wrong. If you look back at the societies of history you will almost always find lawyers, or people offering an equivalent service. I say "almost" because I'm sure there will prove to be an exception, but after a flick through the Wikipedia entries for the major historical civilisations I couldn't find any.

And the idea that laws are complicated because lawyers want them to be complicated is a nice soundbite with no substance. Laws are complicated because the world is complicated. They are an attempt to make clear rules in a world that isn't made up of simple black-and-white issues. They're full of political compromises. And they're drafted by human beings who make mistakes. If the only thing keeping laws complicated was that it was being done deliberately, don't you think we would have fixed that by now?

Yes - bait on an internal network to catch people who see the "shiny" and act.
The question to ask before deploying such things is to ask yourself (or you boss) what your job actually is. Is it to catch a number of people and meet some sort of "arrest quota" or is it to actually protect things? If it's the former then putting up fragile things to attract the attention of the weak willed may be a go, but if it's the latter you may well just be wasting time while the serious threats are getting into your serious systems. They could be getting in while you are distracted playing this game.

IMHO you are better off having better monitoring on the serious systems on a properly segmented network and watching that instead of scattering toys about and looking to see who they distract.

Honeypots are a cool research tool for seeing what people out on the net are trying to do, but as a security measure on internal networks? Sounds more like buzzword overload than anything useful in that situation unless you want some heads on pikes of the entrapped to scare people.

If I'd pulled this shit and enforced some sort of penalty I'd probably be down three or four decent developers because they decided to take a bit of a look around the local network when they first started. Those are just the ones that did really obvious portscans from their own desktop computers so there may have been more.

Very bad analogy. Putting money on the ground in an area with a lot of people and punching anyone that tries to pick it up is a better one.

just pissed because you are a script kiddie

Oh do at least try to grow up.

Instead of putting out bait to encourage people to have a go at fragile systems what about hardening the stuff you've got or put it in segments behind stuff you can harden? Putting out fragile honeypots can lead to wasting time on the merely curious who are no real threat to systems that are not fragile.

The trouble is acting on that knowledge today...

This is David Cameron we are talking about. His previous Big Bold Project involved attempting to build the great firewall of Britain to save the children from internet porn. He isn't one of those people who know how to size up a task...

I'm in the same environment but with clusters. Glueing lots of things together in an interpreted environment is very useful but isn't everything because it usually comes with a speed cost, which while irrelevant in some situations it can consume hours per run in others. Some libraries/tools can take away a lot of the pain (eg. NumPy) but the more your interpreted stuff does the more it becomes clear that seemingly arbitrary blocking conditions are stopping it from doing anything useful for annoyingly long periods of time.
I hit that problem frequently with a dotnet developer who is always complaining about "the network" when his tiny and trivial application runs dog slow instead of having the instant response you'd expect with 1.2MB of data in a CSV. He stopped letting me look at his code after I ran it with the data on the same machine and it was still dog slow (user staring at a blank screen for ten seconds or so, just like with the "network" problem). In that case things have been glued together very badly.

Offtopic a bit maybe, but the big problem of the larger "word processors" is that they try to supply half of a full desktop publishing system as well and it's not the useful half. You can spam pictures all over a page but can't place them precisely - you have to fuck about with other settings and hope they wobble into place.
IMHO it's the feature creep where the word processors approach DTP without getting there that gives that 90% that is marginally useful.
Similarly with IDE's that try to approach LabView but never get there - mainly because somebody points out that a LabView GUI is almost limited to a write-only approach for small and well organised bits of code and an unreadable spaghetti explosion beyond a certain size and complexity.

GP needs to stop playing daddy and let the newbs grow by fixing the problems themselves.

That's when you find users that don't exist in the group files (due to typos) and then when that didn't work a "fix" of setting the permissions of entire systems wide open. Next step some script kiddie has owned the system.
Newbs are often such newbs that they do not know that they should be making their stupid early mistakes on development systems (like we all did) instead of fucking up production systems for everyone - that's one reason to give them a hand even if you are entirely selfish.

Steel cases are still steel cases and can still have sharp edges if poorly designed.

Give me a 10 year old Lian-Li case

That's one of the early examples of the well designed stuff I'm describing. It seemed that around 2000 every new machine demanded a blood sacrifice before it would work :)

Used to use it on Xenix (a LOT) for general text and used it to do a Solaris root password reset about ten years ago but I'd never voluntarily go back if vi is available.

It's because xkcd isn't original content. It just takes existing observations,

Just like political cartoons and a pile of three panel strips. It's what they do.

Gedit is so slow to start that it has been used as a strawman as to "why X is bad" instead of the more accurate "why is this new gnome stuff so much slower than the old gnome stuff?"