Is it possible (and would it even help) to normalize all strings
with regards to surrogate pairs at indexing time instead? or will
container still possibly differ in byte for byte output?
- Donovan
On Sep 11, 2009, at 5:34 PM, Chris Hostetter wrote:
: > why don't we just output the raw bytes ourselves?
:
: That would require writing TextResponseWriter and friends as binary
: writers, right? Or did you have a different way in mind for
injecting
: bytes into the output stream?
Grr.... you're right. i got so turned arround thinking about
counting the bytes and encapsulating it all in the PHPS markup i
forgot
that we still want/need the *raw* bytes output as part of the
character
stream ... not some sort of string representation of the bytes ... wow
that sounds even more comfusing when i look at it on paper.
character data sucks.
I still repeat my earlier vote: let's yank this patch and just
document
that byte counts for strings included in the PHPS format are only
accurate
for characters outside the BMP if the servlet container being used
writes
them correctly -- that seems like a totally fair requirement to
having in
place, and points the figner squarly where in belongs (without
putting us
at risk for having broken code if/when jetty fixes this problem.
Alternately: have an option and/or system property to force/disable
this
behavior even if jetty.home isn't/is set.
Even if we change nothing else: there needs to be a big fat freaking
disclaimer in the javadocs for the writer that it's looking at the
jetty.home variable, and explaining why.
-Hoss
