In the old days, when computers were not so powerful, CAT programs had to index word and phrase lists in order to search them quickly.
With today's powerful computers, this is not necessarily necessary anymore. Glossaries and Memories in plain text have many advantages. You don't have to import or export them and you can edit them directly*) in a text editor. Time savings and comfort.
Wouldn’t it be a good idea to offer a leaner, cleaner, easier to read file format for Memories as an alternative?
Tab-delimited Memories with only the source and target segments’ content in two columns. Okay, add an optional third for tag positions. How simple is that? Smaller Memories, less computer resources required. Better legibility. Nice operations in spreadsheet software possible.
What does Cafetran need TMX for, internally?
*) For example, deduplication, spelling check, global search and replace.