When importing a tab-delimited text file into an MDB, the MDB is first forcibly compacted. This means quickly importing a few small tables (e.g. reviewed translations received from a client) take a disproportionate long time, because many user use a single MDB fro everything, so it is often very large.
This is unacceptable behaviour, a time-consuming operation such as compacting should only happen when requested by the user. The forced compacting does not take place when other formats, like Trados WB, are imported.

Keyboard shortcuts should always have the same effect, regardless of what part of the screen (grid, translation cells, file selector, row selector) currently has the keyboard focus.

Examples:
ctrl-arrowdown (go to next segment),
alt-arrowdown (go to next untranslated segment),
ctrl-h (search & replace),
page down (go to next page in grid),
page up,
etc.

Here are some practical examples of problems with this. @? Ook naar Lexicon.

Ctrl-F (search) should work in addition to ctrl-H (search and replace), both with the same meaning, so it is no longer necessary to learn different habits for MSWord (and many other programs) and DV3.

After invoking the search and replace dialog, the cursor should always be in the field Find What. As in it, this often requires an extra mouse click.

Searching should start at the beginning of the file (or all files, if they are in view) once a new or changed keyword is entered. Currently, the search starts in the middle, even after first opening the project. If the sought word is before the current position, DV3 says it isn't there, without wrapping around to the beginning, while in fact the word is there.

There should be an extra keyboard shortcut: shift-F4 to search for the next occurrence of the current search word. Should work as in MSWord.

In single-file view, temperarily searching in all the files should be possible as an option, without having to switch to all-file view.

When in a search operation the end of the file(s) was reached, DVI should not say "No matches found", but instead offer to continue (wrap around) to the beginning of the file(s). The current message is confusing, because often there are occurrences, which DVI only finds after extra mouse clicks.

If a text search is fruitless in the target language, DVI should automatically also search the source language (and vice versa). This is useful because users tend to make the wrong choice.
(But a search-and-replace in one languages should NOT automatically extend to the other language! Not unless after confirmation by the user.)

A search operation in the MDB (maintenance program) should be interruptable by pressing the escape key. Now it always continues until the end, which can take very long.
Erroneous searches can easily occur if the target language is searched instead of the source language, or vice versa.

In MDB and TDB maintenance, when invoking the search and replace dialog, it should still offer the previous search words and previous source/target choice. In DV 3.0.38 these settings are forgotten each time the search dialog is left (escape key). (The searched item is still in the history list, but invisibly. It should in the text box already.
It is true that there is no need to disable the search box by pressing the escape key, because it is possible to click in the text @? @? (Different behaviour MDB/TDB and DVI? @?

If in a rather sizable project with a sizeble MDB, the options "AutoPropagate" and "AutoSend to memory database" are set, pressing ctrl-arrowdown quickly followed by an alt-arrowleft does not return segment currency to the segment just finished. Apparently the alt-arrowleft is "missed" while DVI is busy writing to the database or searching the project for possible automatic propagates?

In Project configuration, tab Languages:
Changing the target language does not automatically the MDB and TDB to that last ones used for that language combination, this has to be changed manually. If the target language is changed however (one dragged to trash, new language dragged from left to right), setting are automatically changed.
I expect there are more translators with multiple source languages that translators with multiple target languages (most translated into their native language), so isn't very practical.

When the spelling check is interrupted by pressing escape, e.g. to make a correction more complicated than just changing single word, all the words previously involved in an "Ignore all" action are forgotten. They should be remembed during the whole of the spelling checking session.

When the spelling check is interrupted by pressing escape, e.g. to make a correction more complicated than just changing single word, the current segmnent should be the one that was last checked, not the segment after that.

F7 Check spelling should initially (e.g. when pressed for the first time opening the project) should start at the beginning of the file or project (if in all-file view) not somewhere in the middle (current segment).

Deleting an item from the Lexicon should be possible in a consistent manner. Now it sometimes works, but usually doesn't. Also please add the possibility to insert an empty lexicon item

After a ctrl-shift-n (move word under cursor, of current selection, to the right) correctly leaves the selected text selected, so that the selection can be moved further to the right if desired.
Ctrl-shift-p however, for the same operation, but shifting to the left, disabled the selection after being done. A subsequent ctrl-shift-p then move the word under the cursor instead of the selected text. This is so confusing that it makes this potentially very useful function virtually usable.

After sending a marked pair to the TDB using ctrl-F11 or alt-F11, the selection should not be reset, but should stay as it is, so it is easy to add also a slightly modified pair. (Found this to be also a problem in DVX! Then with F11) Voorbeeld: golf course = golfbaan, golf courses = golfbanen. This is useful in language pairs involving German where it can be is convenient to have several case variants in the TDB. Example German to English: symbolisch, symbolische, symbolischem, symbolischen, and symbolischer all correspond to English symbolic. This could be handled with a fuzzy terminology match, but that often finds matches that are too fuzzy.
@? Geldt ook voor Engelse en Franse meervouden.

When exporting from an MDB in Trados WB format, the subject can be exported too. But it cannot be imported from such a file.

In MDB and TDB maintenance, it would be convenient if the last SQL-command used would be remembered so it isn't necessary to enter it again. Workaround: keep often used SQL command in a Word file, and copy-paste from there.

When importing to the MDB or TDB, e.g. from a tab-delimited text file, there should be a possibility to set subject and client for all imported pairs. Now this requires work arounds, like changing ^p to ^tnn^tmm^p in Word (nn = subject, mm = client), or but issueing an SQL-command afterward, selecting on date and time.
The Align function of the MDB maintenance does have subject and client, settable in the menu, but it is easiliy overlooked.

DVI, project configuration, setting the client and subject should follow the normal Windows user interface conventions. Now it often works in very strange ways. E.g. when selecting from the list and pressing Enter, the selected value is not used, but the field becomes empty. The same happens when typing digits (e.g. 62 for Engeneering) and then pressing Enter instead of tab.

In DVI project configuration, the option "ignore hidden text" should be "sticky", that is it has been ticked for one project, it should be on by default in any next project, until reset again.

In DVI project configuration, the default for "Source common directory" and "Output directory for translated files" should not be the path used in the previous project, but the location of the project file itself. In my workflow, this is always the location I use.

In DVI project configuration, the button Browse for the fields "Source common directory" and "Output directory for translated files" should start at the currently entered path, not at the root of the directory tree. I never ever put files in the root, but I often use a location that differs from a previous location only in the right-most portion of the path.

In MDB Maintenance, function Align, the default path for the target file should be the same as the (possibly just modified) path for the source file. Source and target files for aligning are nearly always in the same folder in my workflow. In DV3.0.38 I often have to do the same cumbersome browsing twice.

When creating a new project, the last used target segment of the previous project is often inserted in the first target segment. This is buggy and confusing.

When in MDB Maintenance the cursor is in the source cell, one would expect that pressing the tab key would take you to the target cell right next to it. Instead it goes to the grid, then to an unlabelled field, then to client, to subject and eventually to target.

In DVI and MDB maintenance: pasting into a target cell, after a copy from another application, like Word of a browser, pastes not only the text itself, but also the font size, which is very often much too small. That font then remains active until DVI or MDB maintenance is closed and restarted. It would be better to paste only the text and no formatting information.

When in options "Split text area horizontally" is set (always, in my case) the positioning of the windows for the spelling check is less than optimal: It is just above the target cell, but should be above the suource cell, so that can be see too.
With the option not ticked, both cells are next to each other, so then it makes no difference.

@? Insert mode lijkt soms spontaan uit te gaan, ook al is de inserttoets niet aangeraakt. Misschien is dit een Windows-probleem?

@? Bij in/uitschakelen van Portionsfound e.d. meteen ook de rest van de display resizen.

When enabling ASr (autosearch), so that Sentences found and Portion found appear, the searches should be performed immediately, based on the current segment. This now only happens after going to a different segment. (also a problem in DVX)
Also, it would be helpful is the alternative layouts for DVI (rearrange windows) could them be switched on automatically, so PF en SF can be seen. Likewise, after disabling ASr again, DVI could automatically return to full screen.
This automatic layout switching should probably be optional, because users dual-screen graphics cards may not like it?

Always keep the current segment approximately in the middle of the grid, also after repeated ctrl-arrowdown (for example to acknowlegde exact matches or autopropagates). That way context remains visible, and anticipating for problems in later segments is possible.

Options, clients: it should be possible to sort by client number, so you can see which next higher free number to assign to a new client.

Options, clients: pressing the Enter key after having typed a new client number and client name, should work the same as the Add button, or at least there should be a question if the new client should be saved. Now it silently exist the dialog without saving, as if a Cancel button were pressed.

@? Import in MDB vanuit tab-gescheiden kapt source- en target segment af op 255 tekens! Ontoelaatbaar! Het Trados-formaat doet dat niet. Maar hoe daarin client en subject te zetten? SQL.
Vreemd: bij experiment met kleine MDB is het probleem niet reproduceerbaar! Repair draaien? Nee, oorzaak is waarschijnlijk geweest dat in de cellen (shift-)tabs, shift-returns of returns (in Word zoekbaar als ^t, ^l, ^p) voorkwamen. Word is zo dom om die bij het converteren naar tabsgescheiden gewoon te laten staan, daardoor raakt de hele boel natuurlijk in de war!

@? (auto)propagate werkt niet met dingen als 10/11/05 (veranderd in 10-11-05) of 3,2 (wat 3.2 moest zijn, want hoofdstuknummer). Ook per F12 (propagate to all files) weigert hij het gewoon.

@? F6 (see in context) of codes: should also be in the submenu obtained when right-clicking on the code in question. Now only accessible through the main menu (edit), and as a keyboard short-cut.

@? Ctrl-S terwijl per ongeluk een enkele spatie (of enkele spaties, dus white space) is geslecteerd in source of target: niet daarnaar zoeken, maar handelen alsof er niets was geselecteerd, dus zoek op heel segment.

Code voor vet, cursief e.d. mogen niet gecombineerd worden met een begin of einde van een segment. Voorbeeld: de zin "This was done by XXX"krijgt wel een code voor bold, maar niet voor 'einde bold'. Gevolg is dat de vertaler de zin niet kan ombouwen naar "XXX did that", want dan loopt het bold de hele zin door.

@? Project dat er tijdelijk niet is (wegens encrypted partition die niet gemount is) Niet meteen schrappen uit lijst met recent geschrapte projecten, maar gewoon laten staan.

@? Als de projectfile read-only is, dat melden bij het openen, en niet de verwarrende melding: "Unable to open <pad + file>, or wrong format."

@? Automatisch joinen, als blijkt dat er een fuzzy match is, waarbij het begin van de source volledig matcht en de source van de gevonden fuzzy langer is, en rest klopt met een volgend segment in dezelfde file (zo nodig ook 3 of meer segmenten zo matchen).

DVX: See also: Proposal: AutoJoin.

Often, multiple exact matches are indicated (dark green) although a ctrl-S reveals only a single pair.

In "Count words and characters", do not make "all rows" the default, because such a count is only useful at the start of a project. Instead tick all the other options by default. Or even summarize into a single option, because it hardly ever makes sense to count some types and not others.
As it is now, to know how the project progresses I have to manually check (mouse click or space bar) all those options, each time.

The "Count words and characters" window should be resizeable, so long paths can be seen. Useful when clients use very long names, with the difference near the end.

@? Toevoegen aan TDB (ctrl-F11) neemt soms niet automatisch client en subject van het project over. Waarom eigenlijk niet? Zou vaak wel nuttig zijn, ook al is de term daar lang niet altijd toe beperkt.

@? Opties van de spellingcontrole (bijv. geen woorden met cijfers erin controleren) kunnen bewaren voor de volgende keer.

@? Als er maar één file is, niet vragen "this file" of "all files" (bijv. bij export)

@? A working unlimited Undo. (op z'n minst voor alle wijzigingen sinds het binnengaan van één bepaald segment, zowel voor doel- als brontaal; dat zou al heel veel ellende besparen.
@? Example: just translated a whole segment, didn't press Ctrl-arrowdown yet. Wanted to press F6 to see a strange code, but accidently press F5 (copy source). The whole of the painstakingly composed translation is now lost, Ctrl-Z does nothing. This is unacceptable behaviour.

@? Correct functioning of fuzzies from recently added MDB-entries:
Example: source segment "Removable front panel with grills"
The very next segment is:
Removable front panel
Ctrl-s says "no matches". Several seconds later it does work. Caching problem?

@? Bij omschakeling van weergave één file naar alle files moet de currency gehandhaafd blijven. Nu verspringt dat vaak op een heel onduidelijke manier.

@? De schuifbalk in het rechtervenster (target) met segenten verdwijnt soms.

@? In Portions Found moet niet alleen de linkermuisknop positioneren, waar ook de rechter, om te voorkomen dat een heel ander item wordt verwijderd dan waar op geklikt is.

@? Consistente bediening bij toevoegen en deleten van entries uit het lexicon. Deleten met de Del-toets lukt nu soms wel, maar soms verdwijnen alleen letters. Invoegen is helemaal niet mogelijk.

@? Consistent handling of correction in TDB and MDB in Portions Found. Changes made are now sometimes recorded in the databases, but sometimes they aren't. Sometimes, double clicking twice seems to be necessary to reach an editable field at all.
Oorzaak waarschijnlijk: in TDB staat the=de, in source staat The, Portions Found vindt dat, maakt er correct "De" van (begin van zin), en toont het ook zo. Corrigeren in TDB via PF lijkt geen effect te hebben, omdat het er al goed instaat!

@? An asterisk (*) or number sign (#) at the start of a source segment should not lead to many many seconds of searching (as if the whole MDB is searched sequentially?) when (auto)assembling the translation. An asterisk or number sign should be treated the same as any other normal text.

Show file name in lower right corner of screen, not only in "all files" view, but also when just a single files is in view. This means the user can always look in the same place to see what the current file is.

@?Excel files with lines breaks (made with alt-enter) within cells are not reproduced correctly after export.

@? reassemble soms wel/niet, hinderlijk na een lang segment al gedeeltelijk te hebben vertaald, even terugkijken naar ander segment.

@? The dialog "config project" cannot be left with the escape key, ony with OK. Bij "option" is dat wel zo. Inconsequent.

@? zien uit welk bestand + pad MDB spul geïmporteerd was.

@? Vraag (spaties aan het eind) niet stellen, maar gewoon doen. Of alleen bij Word documenten? Zijn er situaties waarin het niet nuttig of wenselijk is.

@? Als er maar één targettaal is: niet vragen "alle talen".

@? Ctrl-shift-s (zoek terminologie in TDB) geeft wel een scrollbar bij de term, waar die vaak niet nodig is, maar niet bij de omschrijving (of any) zodat die helemaal niet goed in beeld te krijgen is, ook niet bij full screen, want dan wordt het bovenste deel (terminilogie) groot (voorbeeld bij mij: FPSO).

@? Bij melding bij toevoegen aan Lexicon (F9) "entry exists in Lexicon", meteen vermelden wat dan de vertaling was. (Source is al bekend, want gemarkeerd). Zoals het nu is moet je dat zelf gaan opzoeken.

@? Terugschakelen van Lexicon naar bestand(en) waarin je werkte, met Ctrl-F6.

Titel van een document tussen " en ". Had ik eerder vertaald, ik besluit nu op grond van verkregen referentiemateriaal tot een andere vertaling, en corrigeer dit ook in het Lexicon. Dezelfde titel komt ook elders voor. Ik zoek die plek op. Probeer met Shift-Ctrl-S te vinden wat ik net in het Lexicon heb gezet (titel in source geselecteerd), maar hij vindt niets! Ook een terminologiecontrole (Shift-F7, single row) zegt 'Passed', terwijl de term nog in afwijking van het Lexicon vertaald staat.
Dit komt kennelijk door de dubbele aanhalingstekens. Maar dat is een onterechte uitzondering, ook daarin kunnen vertaalbare termen voorkomen.
Maar ook zonder de aanhalingstekens werkt het zoeken niet! Maar Shift-F7 wel.

@? Excel-jobs werken soms niet goed als er newlines in de cellen voorkomen. Voorbeeld W. 20061221

Reverse database: if for the destination database, a new name is specified, that database is automatically created. This is OK. But the language combination for that database is set to that of the original database, so the pair is not reversed. This makes the resulting database unusable, although its contents are OK.
Workaround: first create the destination yourself, with the correct language pair.

Reverse database: if for the destination database, a name is specified that ends in ".mdb" (lowercase), an extension ".MDB" (uppercase) is added to it. This is incorrect, because under Window, file names (including the file name extension), is not supposed to be case-sensitive. So ".mdb" and ".MDB" means the same.

A single word in a segment (in source and target) cannot be added to the Lexicon by F9. The press is silently ignored. There is no reasonable justification for this, the program should simply do what the user tells it to. What's more: this bug applies to any first word in a segment! No, that doesn't explain it either.

Reverse database: if the destination database already contained some data, this is bluntly overwritten without notice! Totally unacceptable behaviour!

Text in text boxes is imported in a direct Word project. However if the Word document is first processed by the Trados 6.5 demo (which does render the textbox texts visible in Word), they do not het imported in a Trados Workbench project. (July 2008).

Sometimes a very simple RTF file, presegmented using the Trados 6.5 demo, doesn't get imported in a Trados Workbench project (using Word97). Example: Job1699, July 2008. Vermoedelijke oorzaak: in dat geval zet Trados 6.5 Demo de tags {0> en <0} niet op de juiste style (opmaakprofiel!), namelijk tw4winMark.

Send project to database, only this file. If the the option "Deleted all pairs prevently sent from this project" is ticked, it really does that, i.e. older pairs from OTHER files in the same project are also deleted. This is nog useful, the deletion should also work per file (and per project).

It is possible to enter newlines in TDB entries and in language pairs in the project files. When exported to tab-separated text, this violates the structure and possibly causes problems when reimporting.