+* Compatibility with older documents/layouts
+
+LyX 1.5.x uses an external python script, lyx2lyx, to import documents
+written using previous versions of LyX. All versions of LyX as far back as
+0.12 are supported, so any klyx users still holding out for an alternative
+to xforms will finally be able to put their dinosaur to rest ;-)
+
+Of course, this means that you must have python (at least version 2.3)
+installed in order to use LyX 1.5.x with your old documents.
+
+lyx2lyx also has the framework in place to be able to convert documents
+to an earlier format (which requires python 2.3.4 at least). However,
+these converters have only been written for the conversion from 1.5.x
+to 1.4.x and 1.3.x, so versions of LyX older than 1.3.0 will NOT be able
+to read documents saved with LyX 1.5.x. The conversion from 1.5.x to
+1.4.x/1.3.x is lossless as long as no new features are used. lyx2lyx
+tries hard to find something equivalent for new features such as boxes,
+but this is known to fail sometimes. LyX 1.4.5.1 contains an updated
+lyx2lyx that can read documents in 1.5.x format. LyX 1.5.x can also
+export to 1.4.x format for document transfer to older 1.4.x releases.
+
+Furthermore, LyX uses a converter layout2layout.py, also written in python
+that will convert old layout files on the fly. You can also call it manually
+on your layout files if you want to convert them to 1.5.x format permanently.
+
+* Preparing for Unicode:
+
+As of version 1.5.0, LyX uses Unicode internally. This is a major change that
+affects documents and layouts likewise. We have tried to do out best to make the
+transition as smooth as possible for you. However, there are some caveats:
+
+- User layout files must be converted to UTF-8
+
+ In previous versions, layout styles were allowed to use non-ASCII names
+ using the local encodings. LyX-1.5 now assumes that all layout files are
+ UTF-8 encoded. This means that non-ASCII style names are still allowed
+ but they must be valid UTF-8 strings. One way of doing the conversion
+ is to use iconv. Using bash, the script below should work:
+
+ #! /bin/sh
+
+ cd /path/to/layouts
+ for l in *
+ do
+ cp "$l" tmp.txt
+ iconv -f latin1 -t utf8 tmp.txt -o "$l"
+ done
+ rm -f tmp.txt
+
+- Inset encodings and Conversion from earlier LyX versions
+
+ As part of the transition to unicode, lyx2lyx (the scripts used for converting
+ back and forth between different versions of the lyx files) converts old .lyx
+ files, which may use a number of different encodings, to UTF-8. This conversion
+ depends on correctly identifying the language of the text. There were previously
+ some edge-cases (insets embedded in different-language text type scenarios) in
+ which the language was incorrectly identified, which caused some text to
+ appear incorrectly after having upgraded from older versions. This has now been
+ fixed. Unfortunately, however, the fix cannot be applied to files which have
+ already been converted past format 249. So if you have already converted
+ your old files (using a development version or release candidate), this fix
+ won't help, unless you still have the originals lying around (and haven't
+ yet made too many changes to the newer versions ;) ).
+
+Generally, it is probably wise to keep a backup of the old version of your
+files, at least until you are sure that the upgrade went smoothly (which it
+almost always will).
+
+* Languages/encodings and insets
+
+One of the bugs fixed in LyX 1.5.0 is that previously, there were certain
+specific cases in which the LaTeX generated did not correctly reflect
+language/encoding transitions in and around insets (footnotes, LyX notes).
+After much deliberation, it was decided not to change older files such that
+they will still reflect the old LaTeX output; rather, they will now correctly
+reflect the situation as it appears in the GUI. This means, however, that if
+you mangled the text in the GUI in the older versions, in order that it
+generate the correct LaTeX output, the LaTeX will now generate the mangled
+text. If this is problematic for you, please get in touch with us on the
+developers mailing list, we do have some possible solutions for this.
+
+The effects of this will be more pronounced for RTL (Hebrew, Arabic, Farsi)
+users -- though they affect users of other languages as well.
+
+* Floatflt in 1.2.x and older