Seems to me (though this thought was better expressed by James Fallows in the now-famous Atlantic article) that the possibility of partial implementation is the greatest thing about GTD. I think my implementation is probably very partial.
At the most basic level you could just start carrying a ziploc folder everywhere to use as an 'In' collection bucket, and process it regularly. Or you could just try to get into the habit of asking 'What's the next action?' All good, even without a bells-and-whistles multiple-context PDA-based thing.
At another level, obviously, it's true that comprehensiveness matters, insofar as the peace-of-mind goal of GTD is best achieved by having *all* your nagging open loops either dealt with or written down in one place, waiting to be dealt with. Half-measures in this sense could be counterproductive, since if some open loops are written down and some aren't, you might be less likely to remember the unwritten ones than if you didn't use GTD at all.
That said I do personally think the initial two-day collection exercise as described in the book is needlessly laborious in one or two ways. Some personal suggestions:
*All that stuff about sticking a Post-It note on each individual bit of stuff, with the next action written on the Post-It, makes things clearer, but if you understand the gist of the process, you could skip it and go straight to making lists.
*As long as you rigorously separate "reference stuff" from "actionable stuff", you *can* hold off from organising the reference stuff properly. I did. And you don't need to buy a labelling machine.
*You could hold off from using contexts completely, if it made things less daunting. Contexts are still the least relevant part of GTD for me, though I know that for some people they are much more fundamental.
*You could decide not to set up a tickler file, and instead rely on your appointments calendar (for essential, non-optional things) and a regularly reviewed someday/maybe file, for non-essential, optional things.
Just my thoughts...