Update client documentation

Improved the client by detailing the failures that can occur on a
client.

docs/client.md

@@ -1,12 +1,32 @@
 # Client Documentation
 
-
 ## Summary
-The clients are their own objects that will request work from the job queue, and perform the work by making calls to [regulations.gov]
-(https://www.regulations.gov/) for data downloads, and saves the results. 
 
-## Description 
-The goal is 
-that the client will request and complete work in order to download data from 
-[regulations.gov](https://www.regulations.gov/).
+Clients are components of the Mirrulations system responsible for downloading data from Regulations.gov. Unless stopped a client continues to attempt the following steps:
+
+1. Get work from the job queue
+2. Perform the job by downloading data from Regulations.gov
+3. Save downloaded regulation data
+
+To accomplish their task each client interacts externally with Regulations.gov and AWS S3. Internally, each client interacts with the database (Redis) and queue (RabbitMQ).
+
+## Details
+
+Every 3.6 seconds a client attempts to get a job, perform its jobs, and save the resulting data.
+Download jobs have three fields: `job_id`, `url`, and `job_type`.
+
+### Getting Work
+
+A client gets work by removing a job from the queue and updating it's current job in the database.
+
+If no work is available at the current time the client waits for 3.6 seconds before attempting again.
+
+### Data Download
+
+After receiving a job, a client attempts to download the remote resource pointed to by its url. If the `job_type` is a comment, any attachments are also downloaded. The client updates the database after the job is completed.
+
+If an unrecoverable error occurs while during download, the client marks the job as an invalid job in the database. Invalid jobs will not be retried by other clients.
+
+### Saving Data
 
+After downloading data it is saved. By default data is saved to disk and to the `mirrulations` AWS S3 bucket.

docs/production.md

@@ -1,7 +1,3 @@
-
-
-
-
 ## Production Environment Documentation
 
 The system is Dockerized into a number of components: