lzText is undocumented. Jan Wieck implemented it, but shortly afterwards had a flash of inspiration and came up with the scheme that is now used in PG 7.1 to implement long tuples (which includes compression).
I (and a couple of others) lobbied the PG crew to leave lzText in, since it was so obviously useful for the OpenACS project. It was decided to leave it in, but as an undocumented feature in PG 7.0 to be superceeded by PG 7.1's long tuples implementation (which works fine with plain-old "text", etc).
"lzText" can be used anywhere "text" can be used and is fully type-compatible with it. Compression/decompression are invisible to the user. In PG 7.1, pg_dump will change "lzText" to "text".
