High-volume delivery is where Exstream's production heritage shows most clearly. The mechanisms below are the historical classic model (HP Exstream 6.1-7.0); current high-volume and postal integrations must be validated per release, but the architecture is worth understanding because most large print estates still run some descendant of it.
Two passes: separate ordering from composition
Output Sorting and Bundling splits the run so customer-order decisions do not have to happen inside final composition.
The pre-sort run processes the application far enough to produce two files: a binary Sort Data File holding processed customer values (never edited), and an editable Sort Index File whose records each carry a 20-character SYS_SortIndex key pointing into the data file.
Third-party postal software can then reorder, subset, delete, duplicate, cleanse, or bundle the index records however the mailing requires, as long as the key values stay intact.
The post-sort run reuses the same package, reads the Sort Data File through the edited index, places late-composed barcodes, images, and fonts, and composes objects configured for post-sort timing.
One rule keeps the whole scheme sound: never repackage between the two runs. Repackaging invalidates the correspondence between index keys and stored customer data, and the failure mode is subtly wrong documents rather than a clean error.
Bundling and householding
Bundling is hierarchical: up to five control variables define nested breaks (region, locality, postal code), and a new bundle begins when a control value changes. Householding is the special case that groups documents sharing an address, cutting mail-handling cost and enabling postage optimization.
The engine exposes bundle state through the SYS_Bundle* variable families (current and total customer, page, and sheet counts), and bundle search keys can be placed before or after a bundle for downstream inserter equipment.
Queues turn ordering into production units
Output queues convert the sorted stream into physical files and machine instructions.
Rules can route documents to different queues by page count or weight, convenience breaks can create one file per customer, event, or bundle, and the queue decides whether a customer document stays in one file or may split across files.
Report files can break alongside the queue, with a variable in their filenames so successive parts do not overwrite each other.
Inserter selection defaults to the first configured inserter unless -SETQUEUEINSERTER chooses another.
Imposition and multiple-up
Multiple-up objects place several pages on a larger sheet; imposition determines the print order of those pages for duplex or booklet production. The historical engine supported 2-up duplex variants with optional back covers and reverse ordering. Z-sort setups required the Sort Index File to alternate customer records before post-sort processing so paired positions landed on the intended sheet sides, a detail that has produced more than one batch of physically misassembled mail when missed.
The contract view
The durable insight from this model: high-volume delivery is a coordinated contract among data ordering, queue breaks, insert control, imposition, and output-driver settings. Resolution, image compression, print-stream optimization, font loading, and reusable resources affect both engine time and downstream device throughput. Change any one element (a new inserter, a postal-rate rule, a different driver) and the contract needs re-verification end to end, on the physical output path, not just in a PDF preview.