mechanics of surprise

I'm interested in mechanics to represent the combat advantage
of surprising one's opponent.  The discussion that follows assumes
a segmented initiative system like Shadowrun or D&D.

We've never found a really satisfactory way to represent the
advantage of surprise.  Some that we've tried:

(1) Side which gets surprise has one free round of actions, after
which initiative is resolved as usual.  In both SR and D&D, at
higher levels this leads to a lot of combats where surprise
completely decides the outcome, often before anyone on the surprised
side gets to act.  This is defensible but too bloody for me.

(2)  Side which gets surprise acts first, then side which lost
surprise.  Behaves more or less like (1) with occasional weird
pathologies (in general it's not good, in segmented initiative
systems, to force all of one side to go "together"--it exposes
weaknesses in the system).

(3)  No mechanical effect of surprise at all.  This can have very
non-intuitive results:  side A completely surprises side B, but then
side B rolls better on initiative and moves first.  But it does
lead to less onesided combats.

(4)  Side which gets surprise acts first, but can only do
"half-actions".  This is what current D&Dv3.5 does.  When normally
a character can move and attack, or move and cast, in a "surprise
round" she can only do one of those.  This reduces the danger
posed by fighters a lot.  Unfortunately it doesn't do the same to
the danger posed by spellcasters.  Also, some characters'
combat manuvers don't work in a surprise round and this can lead
to grumpy comments "we didn't want to get surprise here, it's
screwing us up."  (Of course the characters can wait out the
surprise round without acting, but that feels horribly artificial.)

(5)  Side which gets surprise gets an initiative bonus.  If the
bonus is huge this reduces to (2).  I have no practical experience
with smaller bonuses, but maybe they would be more satisfactory.

For all of these, cases where only some combatants are surprised
should also be considered.  They are perhaps most easily handled by
(5).

Are there any more?  Any comments on picking among these?  I've
tried (1), (2), (3) and (4), and currently use (4) in D&D, but
I'm never really happy with the results.

Mary Kuhner mkkuhner@[EMAIL PROTECTED]