I'm currently working on sort of a DWIM scripting framework for PO files in
Python, that I intend to commit soon to l10n-kde4/scripts, and have just
realized that a split-on-tags script can easily be had on top of it.

After a little tweak to keep apparent widths of CJK characters in sight,
this is what it produces on Yukiko's example with no tags:

  #: kgpglibrary.cpp:281
  msgid ""
  "The image must be a JPEG file. Remember that the image is stored within your "
  "public key.If you use a very large picture, your key will become very large "
  "as well! Keeping the image close to 240x288 is a good size to use."
  msgstr ""
  "画像は JPEG 形式でなければなりません。また画像は公開鍵の中に取り込まれること"
  "に注意してください。大きな画像を使えばそれだけ鍵も大きくなります。最適なサイ"
  "ズは 240x288 です。画像をこれに近づけるのがよいでしょう。"

and upon one of the suitable tags (break after </p>):

  #: keysmanager.cpp:2176
  #, qt-format
  msgid ""
  "<p>Delete <b>SECRET KEY</b> pair <b>%1</b>?</p>"
  "Deleting this key pair means you will never be able to decrypt files "
  "encrypted with this key again."
  msgstr ""
  "<p>秘密鍵のペア <b>%1</b> を削除しますか？</p>"
  "この鍵ペアを削除すると、この鍵で暗号化されたファイルは二度と復号できなくなりま"
  "す。"

However, since the framework parses PO files rather thoroughly, has a lot of
scaffolding for DWIM, and I'm doing Python (as in really coding) since about
ten days ago, the splitter is mighty slow -- roughly 10 times slower than
msgsplit.

That's one reason why it would perhaps indeed be best if the splitting would
be done by setting up save/load hooks in Kaider, rather than by Scripty on
the repository. Another reason is that I would also prefer a suitable form
of --no-wrap, where the lines are split only on \n and the tags, and judging
by this and earlier threads, this does not seem to be a good setup for
everyone.

-- 
Chusslove Illich (Часлав Илић)
Serbian KDE translation team