On 24/07/2015 08:44, Alex Peshkoff wrote: > On 07/24/2015 02:04 PM, Dimitry Sibiryakov wrote: >> 24.07.2015 12:59, Adriano dos Santos Fernandes wrote: >>> There should be no different thing than what we already had with file >>> names. Otherwise, something is completely wrong. >> One different thing: file name is sent to server, password is used on >> client. >> > DB name is also (at least partially) is used by client - to extract host > name from it. > For example searching for "inet:" prefix in unicode text file using it > as utf8 even for english-only script we will have problems. > > We could safely assume that a sequence of ascii characters means the same thing in every charset, as we do not use wide character sets.
For the password hash, it could be done when utf8_dpb is used. I understand we have a problem when utf8_dpb is not used and it means the password text is encoded in the connection charset. I only hope we do not break the architecture completely. Charsets are defined in the INTL library. Using then now, at build time, only for that, would be architecturally worse than declare that passwords could only use ascii characters. Adriano ------------------------------------------------------------------------------ Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
