Here's a question for you to ponder over the weekend: if international military forces were to intervene in Syria, how likely do you think it is that they would be able to create a sustainable solution to the crisis?
The question was asked in an interview this week by the NATO secretary-general Anders Fogh Rasmussen -- and I think it's a good one.
Gideon Rachman, in a thoughtful column in the Financial Times, put it this way: "Is it possible that, in intervening to stop one evil, we will create a greater evil in the future?"
If you or I had been trapped in the hell that was Homs over the past month, I'm sure we would have been begging for foreign intervention, anything to bring to an end a bombardment that has killed countless innocent people, as well as dozens, if not hundreds, of anti-Assad fighters.
But if you or I were an Iraqi who had lived through the hell that followed the US-led invasion of 2003, we may well have cursed the foreign forces that unleashed such horror on our country. The dilemma faced by policy-makers is that stopping one evil does not necessarily protect us from future evils.
Back in 1999, at the height of the conflict in Kosovo, Tony Blair delivered a speech in Chicago in which he outlined five questions that he suggested need to be asked whenever international military intervention is being considered.
1. Are we sure of our case? (Blair commented: "War is an imperfect instrument for righting humanitarian distress, but armed force is sometimes the only means of dealing with dictators.")
2. Have we exhausted all diplomatic options?
3. On the basis of a practical assessment of the situation, are there military operations we can sensibly and prudently undertake?
4. Are we prepared for the long term?
5. Do we have national interests involved?
Syria is not Kosovo, nor is it Iraq or Libya. But if the aim of military intervention is to end the killing of civilians and enable the building of a stable, democratic state, well, let's just say the lessons from Kosovo, Iraq and Libya are far from clear.
As Gideon Rachman pointed out in his column, there is always a risk that by stepping in to prevent people dying, you end up being responsible for even more people dying.
So suppose you simply wanted to offer a basic level of protection to the people of Homs or other cities in Syria where citizens are at risk. How would you go about it? Send in an international ground force, perhaps made up of Saudi, Qatari and Turkish troops? To take on the full might of the Syrian army?
There's been talk of creating humanitarian corridors, safe havens, or buffer zones. But they would all need to be protected by military forces, and one of the lessons from the Balkans conflicts was that even limited interventions can quickly develop into something much more.
The assessment of the NATO secretary-general is that even if there were to be a UN mandate for military intervention (and there's precious little prospect of that in current circumstances), the mission would not have a high likelihood of success.
In other words, it would fail on the third of the Blair tests, even if all the others were satisfactorily answered.
None of this is meant to suggest that military intervention never works. It ended the killing in Bosnia, although it has failed so far to put in place a stable democracy there; ditto in Kosovo (although there is still violence, and countless unresolved issues); ditto in Sierra Leone and East Timor.
As far as Homs is concerned, it may be -- with the "tactical withdrawal" of the anti-Assad forces from the district of Baba Amr and the arrival of the Red Cross and Red Crescent -- that civilians will now at least be spared the constant fear of death by shelling or sniper fire.
What we don't yet know is whether the Syria uprising has reached a turning point: does the defeat of the rebels in Homs mean that government forces can now reassert control across the country, or will the rebels simply regroup and prepare for another stand elsewhere? The northern province of Idlib, close to the Turkish border, may well become the next flashpoint.