Changes between Initial Version and Version 1 of HeaderModularization


Ignore:
Timestamp:
May 27, 2009, 3:30:23 PM (13 years ago)
Author:
Beman Dawes
Comment:

Initial commit

Legend:

Unmodified
Added
Removed
Modified
  • HeaderModularization

    v1 v1  
     1= Header Modularization =
     2
     3== Goals/Objectives/Needs/Wants ==
     4
     5 1. Single include path (-I...). Without a single include path, builds, IDE's, and other tools become unmanageable.
     6 2. Easy checkout, commit, merge, switch, and other source control operations.
     7 3. Works well regardless of build engine; remember that users don't all use bjam or CMake.
     8 4. Doesn't break users existing scripts, IDE setups, etc. OK to break developers and release managers scripts.
     9 5. Easy library subset creation.
     10 6. Minimizes developer mistakes like failure to merge, commit, all files that changed.
     11 7. All maintenance occurs withing the library's libs/... directory. (Subset of (6), but important enough to list separately.)
     12 8. Easy scripting (inspect, unmerged changes, release mgt, etc.) ((7) goes a long way to achieve this.)
     13 9. Works well with installers.
     14 10. Supports test-on-demand and test-library-against-other-branch.