[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ff-cvs] journals tjl

From: Tuomas J. Lukka
Subject: [ff-cvs] journals tjl
Date: Wed, 05 Nov 2003 06:19:25 -0500

CVSROOT:        /cvsroot/fenfire
Module name:    journals
Changes by:     Tuomas J. Lukka <address@hidden>        03/11/05 06:19:25

Modified files:
        .              : tjl 

Log message:


Index: journals/tjl
diff -u journals/tjl:1.167 journals/tjl:1.168
--- journals/tjl:1.167  Mon Nov  3 05:12:58 2003
+++ journals/tjl        Wed Nov  5 06:19:24 2003
@@ -180,28 +180,98 @@
     - possibly a nice performance gain
     - bench first
-- Speed up fenpdf enormously by superlazy papers
 - recursive vobscenes for speedup
     - allows lower priority for buoy nodes
+- Speed up texture loading
+- XXX Should we cache hashcodes in ImageSpans?
 ----- current multitask
 - JGT anisotext
+- pointers
-- Speed up texture loading
+- JoDi rev
+- Speed up fenpdf enormously by superlazy papers
+    - spanimages -> functions
+       - some small API changes towards this; PageSpanLayout
+         constructors &c.
+       - start making tests for FenPDF to make sure
+         I don't break anything
+           - fix run_fenpdfdemo by removing remaining
+             ppActions in create demo space
+           - reorganize fenpdf10.py and fenpdf.py a lot,
+             documenting and removing crud; now most of the
+             creation happens inside the FenPDF object itself.
+             Most parts are still easily replaceable,
+             those that aren't need to be fixed
+           - make first test that checks that FenPDF construction
+             works
+    - pointers
+    - thinking about how to not lose interactivity during
+      PDF texture generation
+    - thinking about spanimages as functions, to allow
+      superlazy papers
+       - main problem: caching of imagespan -> singleImage,
+         as several imagespans map to the same image.
+         Currently taken into account in a special cache.
+         Probably requires a change in Alph to allow
+         this to work without creating tmp objects.
+       - make Alph cache the single-page objects
+       - accept and implement imagespan_getwholepage for
+         the same purpose
+    - document PageScroll better
+       - change accidental Span return type to PageSpan.
+    - heard through Benja from JoDi: rewrite required, but
+      accepted given that. Yippee! OTOH, now I'm stressing
+      a lot about doing the rewrite correctly. Allegations
+      of technophobia and scalability need to be answered, indeed...
+      For some people, it *is* true that I may appear technophobic,
+      as I go against the mainstream current by not believing
+      in making computers do "magic" stuff
+      like understanding what a person is trying to do
+      or understanding the commonalities between different
+      data formats without human intervention. Yes, computers
+      can do 50-90% of the work there, but for many things
+      like that to make sense, it should be 99% or more.
+      Google can find documents but only if the user first
+      thinks about what the keywords &c are. Finding document
+      *related* to a particular one is a FAR more difficult problem;
+      it would require the computer to distil the information like
+      the human does, but the computer doesn't know what the words mean.
+      In short, like Nelson, I do believe that computers are
+      better doing things that they're told; the interesting
+      problem is that there are several interesting things
+      that current computer systems don't *want* to be told,
+      like transclusions or bidi links. AI is an interesting 
+      field by itself, but applying AI to UI yields Clippy all
+      too easily. Sometimes, it makes me wonder whether even
+      some *researchers* are subconsciously anthropomorphizing
+      computers...
     - synchronous loading of indirect mzms for benching fenfire
     - small speedup of ff nvblur papermaker
     - thinking about adding virtual 4x texture SS to fenpdf, as
       well as speeding up nvblur by not using TXD
+    - pointers
 2003-11-02: spending time with the family
     - figuring pointers structure w/ benja

reply via email to

[Prev in Thread] Current Thread [Next in Thread]