In order for the prefetching algorithm to work correctly, it needs a
precise chronological history of each session. Prior to our
modifications, the Mosaic global history only kept track of the URLs
visited. The data structures were set up to store last access timing
information, but this was never implemented. Our code expanded on the
original data structures by adding working fields for access times,
access counts and size of the accessed document. In addition, a
chronological history was implemented as a linked list, with each node
containing the following information: URL, access time, duration of
network access, as well as an inline image flag. Unlike the original
global history, it was necessary to include all inline images as
separate entries in the chronological history. This is so the inlines
can be included for consideration when the prefetch tables are
created.
The history can be saved to disk in one of two formats: a
slightly modified version of the original Mosaic global history format
containing URL, last access date, total access count and size, or our
strict chronological format. This format contains URL, access date,
access duration, inline image flag and size. Note that the
chronological history may contain multiple occurrences of a given URL,
since each and every access is recorded (including inlines, as well as
data found in memory or disk cache). An example of the chronological
format can be found in Back |
Index |
Forward