daniel123 - already sent you a response on this by e-mail but realized I should probably have posted it here instead, so here it is:
There are two different ways a card can contain a user-created definition; the card itself can be custom, in which case the "custom" option is selected in Edit Card and the definition is stored right in the flashcard database, or the card can link to a user dictionary entry, in which case the card is actually fixed / non-editable and the definition only changes and you go in and edit the user dictionary entry it links to.
So it sounds like what you were doing was taking cards that linked to user dictionary entries, converting them to custom cards (in which case the definition gets duplicated in the flashcard database, so you have two different copies of it), changing the definitions on those cards, and saving them. That actually should change the definition on the card - if you go back and examine the card again after changing the definition, has the card changed? It wouldn't affect the user dictionary entry, though - you'd have to edit that through the "Edit Entry" screen.
Importing your cards with the user dictionary checkbox unchecked would store all of your custom definitions in the flashcard database, so there wouldn't be any mixup with cards linking to user dictionary entries. And it wouldn't prevent you from creating other user dictionary entries, it would just keep your user dictionary cleaner / clearer by not putting all of your custom flashcards in it.
After that new import, you can still convert the flashcards that you do want turned into user dictionary entries (i.e. the ones that aren't sentences, I imagine) using the "To User" command in the Manage Flashcards "More Options" screen. Do an "Advanced" search for cards with a length less than or equal to 4 (for example), then use More Options to batch convert all of those to user dictionary entries. You can also do this one-card-at-a-time with the "To User" button in the Edit Card screen.
radioman - there really isn't much a user can do with regards to memory management on Palm, just try not to have too many applications running in the background (MP3 players etc) while Pleco is running. We found a really nasty heap corruption bug last night that could actually explain almost all of the instability-after-a-big-Palm-operation issues, though; didn't actually involve the system running out of memory, but statistically it was much more likely to screw up something important on a system without much space memory than on a system with plenty of it. (covers the crash-on-exit issues, too, ipsi/Fleminator/et al; it turned out the system was crashing after Pleco had already finished unloading, so about the only explanation for that was heap corruption, and having finally managed to reproduce the problem here we've now un-reproduced it with this fix)
I see what you mean now about calendar time - that should be easy to fix, we'll just check for that today button and have it reset both boxes. Resizing that category popup list is actually kind of tricky because of the way the software's designed - the part of the system that lays out that form doesn't know what buttons are going to go in which places and so it would take a bit of redesigning to get it to figure out / resize the popup list for the appropriate button to make it larger.
llammamama - thanks for finding that, but I'm still not sure if setting the repetition interval so potentially low is such a good idea - with SuperMemo they can get away with it since they track a lot more information on people's study habits / how well they retain words they've seen after such-and-such interval than we do. Plus, their newer algorithms calculate the next repetition interval using a whole fancy matrix-based system, whereas we just multiply by the difficulty factor, so lowering that initial interval would have a much bigger effect on the card's later spacing than it does in SuperMemo.
With the default aggressiveness setting, for example, the minimum difficulty score (= most difficult, we probably should try to make that less confusing) a card can have is 60, so the score after the first correct answer would be just 240 (= 2 days); it would then grow by just 20% (the difficulty score only going up if you answered it completely correctly, i.e. with a score of 6, and then only by 4 points each time) each time you got it right, so it would take two more correct answers just to get the interval up to 3 days.
With the minimum aggressiveness level it gets downright silly, that can be as little as a 10% increase (minimum difficulty of 74 but the divisor's bigger - this is another problem, the minimum difficulty is higher with the less aggressive algorithms because of that bigger divisor), and the default score after the first correct answer is just 200, meaning the first score would be just 148 and it would take 4 correct answers just to get it up to two days. (of course it would take 5 correct answers to go from 2 to 3 even if the initial score was 200, but that's already bad enough)
goog1e - there already are 5th tone recordings for a lot of syllables, actually, at least the ones that are commonly heard in 5th tone like de and ma. We didn't design the recordings to be used for syllable-by-syllable readings, really, we just put that in because we realized people would get annoyed if nothing came out when they hit the audio button - once we release the full set of audio files, almost all common words will have dedicated multi-syllable recordings that will sound a lot better than anything we could do automatically stitching together individual syllable sounds.
And that's also why applying appropriate tone changes to 不/一/etc isn't something we're likely to do for now - we may eventually try to add a "real" text-to-speech type system, in which case we'd do new recordings of single syllables with each of them in several different pitches / durations (a mid-word 不 is supposed to be a lot shorter, for example) and stitch them together more intelligently, but our current set of recordings aren't designed for that and couldn't really be adapted to do that well. And if we start transforming 不/一/etc on the fly people are going to assume we're doing other corrections dynamically too - better to have it clearly sound awkward / non-standard in that syllable-by-syllable mode so people don't mistakenly think that this is a correct pronunciation for the word.
With tone-practice / free-answer I think there's actually a pretty good argument for requiring people to enter the original, non-sandhi tones - forces you to learn about those tone changes, which you need to do to hear / pronounce Mandarin correctly. Plus, entering the sandhi tones keeps you from learning the correct, official tone for each individual syllable, which is really the main goal - if we accept ni2hao3 as a correct answer along with ni3hao3, you might think 你 is supposed to be second-tone, or at least have the mental association muddled in a way it wouldn't be if you always had to enter it as ni3.
On the flashcard bugs - good catch on #1, there are a couple of glitches in backtrack mode actually (for example the category toggle button icon isn't updated, though pressing it does alter the category for the currently-displayed card). We've been dealing with a bunch of bugs like #2, found / fixed the same issue with entry lists in the last beta. #3 is probably an unannounced crash, actually - there's a nasty bug (fixed for B7) which can either screw up the first card of an incorrect-card review or cause it to crash after the last card before that review begins. Same with #4, I think - if the problem persists in B7 definitely let us know. #5 we've seen but are only reproducing intermittently, we'll hammer away at those modes specifically and see if that helps us find it more consistently.
benzhen - sounds like the software's treating links in the definition like links in the original text (which aren't even really supported yet themselves) - should be easy to fix. The buttons might optionally go away in 2.0.x/2.1 but it has to be optional and we're really trying hard to stick to our no-new-checkboxes policy until we get 2.0 out the door. Glad the position remembering is working. With the GB bug, the problem there might be that it's GBK or some other expanded version of GB that supports traditional characters - we should probably add support for those too eventually. That Chinese subdirectory search issue is indeed a WM bug, all we do is tell it we want a dialog showing all .txt files and it goes through and searches / builds the file list itself - not much we can do about that except to roll our own dialog (maybe at some point but not for 2.0).
