Juri Linkov wrote:
>> Well, if it can't be done, then maybe TAB should be allowed, but only if
>> the user turns it on (globally, not per input method, since multiple
>> input methods will need to use it).  It should be off by default.  That
>> way TAB is still useful with Quail to do indentation, but if the user
>> knows he won't need TAB when in a Quail input method, he can customize
>> it to do `quail-completion'.  That seems like a decent compromise to me,
>> what do you think?
> 
> I agree but currently have no idea about implementation.

I am not sure I understand the details here (or rather "I am sure I do 
not understand all the details here"), but things like this is what 
tabkey2.el tries to solve.

I have done some significant changes since I last proposed it. I have 
attached a new version.

Could you please see if what you want can fit in tabkey2.el? (The idea 
there is that normally the first Tab does indentation while the second 
Tab key can do some kind of completion - and S-Tab can choose what to 
complete.)