Understanding-mathjax-performance.md

@@ -124,11 +124,13 @@ We should investigate how to optimize this. Some ideas are
 
 One way forward is seek new ways to optimize the SVG and HTML output.
 
-As mentioned, the SVG output is often 25% faster than the HTML output. Improvements could only be made if browsers themselves become more reliable.
+As mentioned, the SVG output is often 25% faster than the HTML output. The HTML output could catch up to SVG if certain measurements could be dropped. But this can only succeed if browsers themselves become more reliable.
+
+Another idea is to re-use formulas or even subexpressions, i.e., if an equation appears multiple times in a page, we could try to only render it once. It's not clear how much of an advantage this is. Most formulas do not appear the exact same way over and over again. Small changes in CSS (size, positioning) could damage the quality when re-using.
 
 We can investigate current javascript optimization techniques.
 
-We can also develop speed profiling tools for content providers to narrow down performance problems related to MathJax.
+We can also develop speed profiling tools for content providers to narrow down performance problems related to MathJax on individual sites.
 
 
 #### Optimizing perceived performance