mirror of
https://github.com/samsonjs/vdirsyncer.git
synced 2026-03-25 08:55:50 +00:00
Add rationale
This commit is contained in:
parent
85bbcc718f
commit
2518a7231a
1 changed files with 20 additions and 0 deletions
|
|
@ -61,3 +61,23 @@ Reading from vdirs
|
|||
|
||||
- The ``ident`` part of the filename *should not* be parsed to improve the
|
||||
speed of item lookup.
|
||||
|
||||
Rationale
|
||||
=========
|
||||
|
||||
Performance
|
||||
-----------
|
||||
|
||||
Currently, vdirs suffer from a rather major performance problem, one which
|
||||
current implementations try to mitigate by building up indices of the
|
||||
collections for faster search and lookup.
|
||||
|
||||
The reason items' filenames don't contain any extra information is simple: The
|
||||
solutions presented induced duplication of data, where one duplicate might
|
||||
become out of date because of bad implementations. As it stands right now, a
|
||||
index format could be formalized separately though.
|
||||
|
||||
vdirsyncer doesn't really have to bother about efficient item lookup, because
|
||||
its synchronization algorithm needs to fetch the whole list of items anyway.
|
||||
Detecting changes is easily implemented by checking the files' modification
|
||||
time.
|
||||
|
|
|
|||
Loading…
Reference in a new issue