This thread is pretty old but here is my $0.02 worth ..
The secret to using an end-fed wire is to pick a length that will present a reasonable impedance on the bands of interest so that it is possible to match the wire with a tuner. The key to this is to avoid lengths that are multiples of a half wavelength on the amateur bands, as this will present a very high impedance that will be difficult for many tuners to match. Fortunately someone has already figured out what these magical lengths are.
Alan Chester G3CCB (SK) wrote an excellent article called
"Taming the End-Fed Wire" ("The Antenna File", RGSB, pg
118) he looked at this issue by plotting various 1/2 wave lengths and
proposed the length of 26.5m (86.9 ft) for 160 - 10M use (there are a couple of other shorter lengths as well that exclude 80m and 169m)
(Note that "The Antenna File" book may still be available from ARRL).
I can personally vouch for the 26.5m length as I currently use this length for an inverted-L (30 feet vertical and the rest horizontal).
As others have said, an inverted L requires radials (the more the better but if you can't put down more than about 20 then don't bother to make them any longer than 1/8 wave on the lowest operating frequency as the performance improvement will be very small).
My 26.5m wire comes in the window of the shack and is connected to a
SG-211 autotuner (this has a fairly wide matching range) It will load from 160m to 10m with this tuner and antenna performance is reasonable. I have put down as many radials as I can on my small city lot and made a good connection between the ground terminal on the tuner and the cold water pipe a few feet below the tuner in the basement.
Another option is the W3EDP, which coincidentally uses a length of 84 feed along with an elevated "counterpoise" (I hate to use that word as it meaningless) of 17 feet (google W3EDP for more details). http://www.zerobeat.net/g3ycc/w3edp.htm
has some pretty good and accurate info on the W3EDP.
W3EDP arrived at this length by starting with 100 feet of wire and trimming 4 feet at a time and then assessing the usefulness of the antenna, he repeated this several times and settled on 84 feet. I think that this is the same 26.5m that G3CCB came up with, only arrived at by experimental rather than theoretical means. Note that the counterpoise length of 6.5 feet is proposed for 20m and I believe that that original QST article on this antenna suggested that no "counterpoise" be used for 80m and 10m.
So for what it is worth, if you decide to go with a multi-band end-fed wire then I suggest that you make it 26.5m long, as this will facilitate matching. This certainly won't rival stacked mono-band yagi's at 100 feet but it will get you on the air on all or most bands with a reasonable tuner. There may be some particular quirks with the radiation patterns on some bands but at least you will be radiating a useful signal.