Let's call it the Westminster Paradox, shall we? I can't believe it has always been true, but it is definitely true now.
It is this. Westminster politicians constrain themselves in a whole range of ways in their understanding of what they can change - they are constrained by international trade treaties and treaty obligations, by a creaking economic belief in 'trickle down', and the overwhelming need to carry on with the economic consensus of the 1990s. It sometimes seems as if they dare change very little.
On the other hand, get them on a conference platform, get them talking about the NHS and they sound staggeringly and unrealistically powerful - 'grandiose', the psychologists would call it.
Ed Miliband promises the finance for another 30,000 more NHS staff, unaware perhaps that the finance is the least of the difficulties here - you can't just conjure up 30,000 professionals in a year or so. Are they in training? Are they already trained and wanting to hear his bugle call? Or will his people go out to the developing countries and offer enough money to their newly trained professionals and ship them over?
David Cameron promises seven-day-a-week GP surgeries, at the same time as his Chancellor promises continued austerity - at a time when primary care is shuddering under the impact of extra costs passed on by private NHS contractors and the peculiar by-product of contract and target culture and payment-by-results. Where will these new GP surgeries emerge from? Where will the extra doctors come from? Are they in training?
The answer is that they are not, at least on that scale.
But then Cameron has realised - as we all have - that the issue of getting an appointment with your doctor is going to be one of those key election issues that can sink a sitting government.
It has a symbolic value for middle England, despite the fact that - in practice - many of the most imaginative practices have managed to find solutions.
In the Blair years, this would have been solved with a vacuous piece of sticking plaster - a 48 hour target, which gave people the right to see a GP in two days if they needed to. Targets always have perverse ways of making the situation worse and this one did so especially - soon you could only get an appointment within 48 hours, no earlier and no later, and practices hoarded their appointments in the most bizarre and irritating ways.
But we have a different kind of target these days, called something else. Also, there is no doubt that surgeries need to expand their horizons, and to take back responsibility for the disastrous out-of-hours care, which was taken away from them in the equally disastrous pay agreement under the Blair government in 2004.
But then, where is the money to come from? Contracting out the out-of-hours service has been so disastrous that handing it back to GPs would cost a small fortune, just in increased insurance costs.
We also need to know rather better whether demand is actually rising in primary care and why. My own sense is that this is, at least partly, the result of target and contract culture.
This has tended to go hand in hand with contracting out to the private sector, but it actually has no necessary connection. The problem isn't which sector delivers healthcare, it is what happens inevitably if you try to define the numerical outcomes which a contractor is responsible for delivering, and to squeeze the cost at the same time.
By chopping deliverables up into figures that are easy to measure and report on, all the rest gets lost - and the resulting costs land on the NHS as a whole. Staff find themselves under pressure to minimise their broad efforts, except where it relates to crossing the numerical thresholds.
See my book The Tyranny of Numbers about the perils of too much measurement.
None of this would matter if you could actually measure the full range and depth of what a good health professional does, but you can't. Anecdotal evidence suggests that doctors who build a relationship with patients deal better with risk - the patients need less reassurance and the doctors commission fewer tests.
Actually, nobody as far as I'm aware has tested this hypothesis - mainly, I suppose, because it flies in the face of current assumptions.
So there's another paradox. The more you focus on narrow costs and numerical deliverables, the more costs go up.