Look at
http://www.microchip.com/Download/appnote/category/rdesigns/40171a.pdf
it should answer any of your questions.

> At 01:28 PM 8/22/01 -0500, you wrote:
> >Had some trouble with what seemed like a simple application:  using a
> >12C509A to control a triac that passed generator AC to a light bulb.  I'm
> >the software guy, so I apologize if this description is not adequate.
> >Two of the pic pins were connected thru 270 ohm resistors to the triac
> >(a Teccor Q2008L4).  The pins were driven low for 150us to turn the triac
> >on.  Two pins were used to ensure adequate drive current for the triac.
> >The pic was very inconsistent - it seemed to run properly some of the
> >time, but had problems most of the time.  The hardware guy said that the
> >way this circuit was arranged, Vdd + 1 was present on the load side of
> >the 270 ohm resistors whenever the triac was on.  Since this would feed
> >current back into the pins, and the 1V exceeded the device spec, and the
> >pic was behaving strangely, he added a transistor between the pic and the
> >triac (so only one pic pin to turn on the transistor is required now).
> >I used a fresh chip for the software, and now everything runs fine.  My
> >questions are:
> >- what deleterious effect was the original design having on the chip?
> >- was that likely the cause of the strange behavior?
>
> Putting current back through protection diodes, it could cause latch-up
> or other bad effects. A little-known fact is that triacs have a voltage
> on the gate (when on) that is the same polarity as the load current.
>
> >- if this was doing something bad to the chip, are the ones I used before
> >  (which were windowed parts) likely to be permanently damaged?  Is there
> >  a way to test them?

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.