Documentation Center

Segmentation Caching in WorldServer

When an asset has been segmented, WorldServer stores the segmented content in a cache. Future processes that need to access the segmented asset will first check the cache to see if there is an existing version in the cache that can be used. The segmented asset cache depends on several items:
  • The source asset fingerprint. In most cases, the fingerprint is equivalent to a file-last-modified timestamp. In other cases, such as the SQL connector, it is a checksum of the column content.
  • The file type associated with this asset
  • The file type configuration
The cache will be dropped and the asset will be re-segmented automatically only if the asset, the file type, or the file type configuration has changed since the segmentation was last performed. If an asset or file type has not changed since the asset was last translated using a workflow, the new project will use the same segmented asset structure that the previous project had, and new TM entries will not be re-applied.
Caching is important for several reasons:
  • Segmenting is expensive because it involves populating database tables with segments and looking up Translation Memory (TM) matches.
  • In ad hoc mode, you can close the Browser Workbench and later return to the work you were doing. Caching the asset preserves the work; if the asset had to be re-segmented on reopen, then the work would be erased.