The code has certainly improved since last time. Now glyph names are
being used directly, without any round-trip to Unicode code point and
back to character code. This is great as it simplifies the code quite
a bit and makes it more robust.

The PostScript parser hasn’t changed so my concerns about its resilience
against ill-formed fonts remain. Also, while memory usage has improved
a bit, there still seems to be unwarranted copying of byte arrays here
and there. For example, the creation of the encrypted portion could be
made on the fly by wrapping the output stream into a FilterOutputStream.

Overall there is room for further streamlining and simplifying the code.
But since it’s an optional feature I suppose there is no harm in letting
interested users experiment with it. So I’ll vote +1.


On 14/05/14 10:22, Robert Meyer wrote:
Hi All,

Following on from the last failed vote for adding Type 1 subset support, I have 
now put forward a modified patch and am ready to try this again. The patch went 
up on Monday to address the issues and comments made by Vincent and Luis. This 
vote will last 5 working days and will finish next Wednesday at the same time.

As always if you have any concerns or find anything please let me know. If they are small 
I will try and address the issue before the vote is finalized in order to avoid repeating 
this again. Please note that all type 1 fonts will default to full embedding unless you 
use the embedding-mode="subset" as per one of the recommendations from the last 

Here is my vote: +1

Best Regards,

Robert Meyer

Reply via email to