3. CAPS™ communication infrastructure

CAPS™' layers 1, 2 and 4 communicate securely among themselves with the SSL (Secure Sockets Layer) protocol, as implemented by the OpenSSL library. Python's interface to OpenSSL, known as PyOpenSSL, was developed by Strakt and released under the GNU Lesser General Public License (LGPL). [Layer 3 runs within the same process as Layer 2, and therefore intrinsically communicates with it securely and rapidly by typical intra-process mechanisms such as function calls and callbacks].

SSL runs on top of TCP/IP. Therefore, the processes that run the programs composing CAPS™' layers 1, 2 and 4 can (and, in fact, generally do) run on multiple hosts in the same LAN. This "naturally distributed" architecture eases the deployment of scalable systems, with optional load sharing and redundancy.

Secure communication among CAPS™ layers 1, 2 and 4 is also quite feasible over WANs, and indeed even over the Internet. However, performance issues related to network traffic (bandwidth and delay) need to be carefully assessed in order to make such deployments practical. As a rule of thumb, most front-end (layer 4) programs may often be successfully deployed at locations remote from upstream layers. Deployment of server-layer programs at locations remote from each other, on the other hand, normally requires reliable, high-speed connections.

CAPS™' communication with other systems employs various appropriate protocols and formats. CAPS™ uses SQL (and modules compliant with Python's DB/API specifications) to interoperate with relational database systems (including the "layer 0" database). SMTP and RFC 2822 formats (as implemented by Python libraries and extensions) let CAPS™ interoperate with e-mail systems. Other standard formats and protocols, such as HTTP, HTML, and XML, similarly allow CAPS™ to interoperate with Web servers and browsers, Web services, and other applications yet.

All of the communications-related components within CAPS™ use a highly scalable and responsive asynchronous architecture. This architecture employs both event-driven processing and multi-threading, and is implemented on top of the "Twisted Matrix" Python communications framework.