The common header specifies the length of each record. Consequently, it is possible to parse the Escher record stream without knowledge of the actual contents of each record. The Escher team intends to take advantage of this fact in future versions. As new features are added, Escher will define new record types. Readers of the Escher file format should skip over record types unknown to the reader. In addition, readers should not expect a record to come in a certain order in a container. They can, however, expect that the containership hierarchy will not change. For example, readers do not need to handle the case of a shape record containing a drawing record.
When Escher writes to a client file, it stores client-specific records in it stream to preserve the client features and behaviors. On the other hand, when Escher writes to a clipboard stream, it uses a client-independent form of the file format to allow interchange between applications.
Escher saves records in Intel byte-order even on the Macintosh. The Macintosh version of Escher byte-swaps the records as they are loaded and as they are saved. Records are tightly packed, without alignment. The LONG type is 32 bits in length.