unbraced length of the compression portion of a long member

Generally speaking the main application I can think of for trusses with unbraced compression chords is pony truss bridges. The AASHTO (ugh) Guide Spec for pedestrian bridges might have some information. I also believe the American Wood Council's NDS might have a paragraph with prescriptive requirements for unbraced wood truss chords in compression. It really depends on the material and what code you're designing to. I think the general idea of anything you find, though, will be that you need to either:

1. Go with the conservative full-length assumption.
2. Brace it.
3. Mathematically demonstrate according to some accepted methodology that the truss web members and connections are torsionally strong and stiff enough to brace the chord internally (i.e. looking at the truss globally as a single beam and checking it for lateral-torsional buckling).
4. Do a fancy advanced second-order analysis that explicitly models all system imperfections, any reductions in stiffness due to partial yielding, etc. so you can just throw the concept of effective length out the window entirely (most common for crazy curved structures drawn up by architects that defy any conceptual simplifications like "effective length").

I definitely agree that a sketch is needed, though. For example, it might help demonstrate why the ends of the bottom chord are in compression. Without a picture, right now the two possibilities I'm thinking of are:

1. The truss is restrained at both its top and bottom chords and is behaving as a fixed-fixed beam.
2. The ends of the bottom chord are only in compression in certain load cases due to some sort of applied axial load, like a braking crane, but the whole chord is in tension in the absence of such loads.

In scenario 2, it MIGHT be justifiable to treat the compression portion by itself as an axially-loaded, cantilever post.