Sidebar: File Differences

The most visible differences in the mainframe-hosted versions of the internet applications are products of the significant difference between UNIX and IBM mainframe files. In the mainframe environment, the filesystem comprises "datasets," not "files." There are four kinds of datasets on the mainframe: sequential datasets, partitioned datasets, indexed sequential datasets, and direct access datasets (see the sidebar "IBM Datasets" for details). (For clarity, throughout this article, when dataset is mentioned, I mean a mainframe file. If I am referring to a UNIX file, I simply call it a file.)

Although UNIX filespace is dynamically allocated as needed, space for mainframe datasets must be explicitly allocated before the dataset can be used. Thus, to write into a dataset, an application like ftp must be able to estimate the amount of track or cylinder space ahead of time. Should the dataset fill up, it is then put off to tape, or some other secondary storage device, reallocated with more space, and then restored. To accommodate this difference, the Interlink adaptation of ftp has been extended with special site commands. Mainframe ftp users must use the site command to create an appropriate dataset before initiating each file transfer.