One consolation is that Salesforce is built with many products I touched during my open source decade. I worked with Doug Cutter and Jon Stevens to migrate Lucene to Apache Jakarta. Salesforce uses several Apache Commons products that I helped create (along with the Commons itself). So, although I don’t get to work on open source projects as much as I used to, I am grateful to be working with a solid, thoughtful product built on a backbone of truly awesome open source components

Core infrastructure

Salesforce.com pioneered cloud computing. Its combined Software-as-a-Service and Platform-as-a-Service offering handles 100,000+ customers, over 2.1 million users, 30+ million lines of third-party code, and hundreds of terabytes of data, running on its own blend of open source, proprietary, and custom code. [Wikipedia]

While the Salesforce platform uses the Oracle Database system as the basis as its backend data store, the web infrastructure is powered by several open source products, like Caucho Resin. The Resin Application Server is a high-performance XML application server for use with JSPs, servlets, JavaBeans, XML, and a host of other technologies. [Hurwitz]

Fulltext search

To help users find data quickly and easily, in 2002, Salesforce.com added the open source Apache Lucene search engine to its technology stack. Today, Lucene to manages an eight terabyte index of about twenty billion documents on the Salesforce platform. The Salesforce/Lucene cluster consists of roughly sixteen machines, which in turn contain many small (sharded) Lucene indexes. The system handles roughly 4000 queries per second, and it provides an incremental indexing model where the new Salesforce user data is searchable within approximately three minutes, through use of a custom optimizer. [Igvita]

Quality assurance

Salesforce.com brings out three major releases a year, and to keep development on track, their quality assurance team runs 50,000 tests a day. About 20,000 of those tests run under the open source Selenium web application testing system. Selenium supports exporting tests to native application languages, like Java and C#, and a Salesforce.com engineer open sourced an extension for its own language, Apex. [Porro, Nagata]

User-facing tools

Several of the Salesforce external data manipulation tools are built with open source components or are themselves open source. The indispensible Salesforce IDE is an Eclipse plugin. The popular Force.com Excel Connector is hosted at code.google.com. Responses to Outbound Messages are handled by the Apache Commons Connection Pool. The very useful Apex Data Loader is powered by the Spring Framework and Apache Commons BeanUtils. Since the Data Loader is open source, an OSX version sprang up (LexiLoader), and there is also an open source wizard (DataLoaderCliq) to help with Data Loader automation.

Advocacy

At its 2011 Dreamforce Conference, Salesforce.com offered an entire track of presentations that encouraged users to adopt other open source tools, including a developer session entitled “Build Better Apps with Open Source”. On an ongoing basis, Salesforce.com hosts an Open Source forum as part of its support site. [SFDC]

By leveraging a solid core of open source packages, and mixing in its own proprietary customizations, Salesforce is able to deliver a robust, configurable platform, upon which millions of business users rely every day.

Products mentioned

• Salesforce CRM
• Caucho Resin
• Apache Lucene
• Selenium
• Eclipse
• Force.com Excel Connector
• Apache Commons Database Connection Pool
• Apex Dataloader
• Apache Commons BeanUtils
• Spring Framework
• LexiLoader
• DataLoaderCliq

References

1. [Hurwitz] Is there beef behind SalesForce.Com?
2. [Wikipedia] Salesforce.com.
3. [Igvita] Lucene in the wild: Salesforce, LinkedIn, Twitter, et al.
4. [Porro, Nagata] How to test Salesforce Apps with Selenium.
5. [SFDC] Apex Data Loader, Build Better Apps Using Open Source, Open Source Labs tract, Open Source Forum.

The Internet was designed to survive a nuclear holocaust, but no one expected to find social media networks still available after the Sun’s surprise flare-up on Saturday.

In a previous blog, we walked through example code for an Outbound Message Listener. We covered the essentials, but Salesforce CRM offers a nifty twist on just sending an outbound message to an external system. Not only can a system receive a message, it can send back a response. For example, if the outbound message inserts a record, the external system can send back the record’s key.

Our example Listener called a worker method that handled the messages in a foreach loop. Using the case of returning an external key, we can add the keys to a hashtable, and send back a batch at the end all of the responses. We’ll pickup the live action towards the end of the foreach loop from the previous example.

 Hashtable responsePayload = newHashtable();  // Returns IDs to SF
 foreach (AccountNotification message in n.Notification)
 {
     sf = message.sObject;

     // – Check cache –
     // …
     // – handle message –
     // …
     // Queue ID for return to SF
     responsePayload.Add(sf.Id, my.Id);
 } // end foreach loop
 // – Return IDs to SF –
 log.Info(“Info: Sending response to Salesforce …”);
 Response response = new Response();

 log.Info(response.Send(“Account”, responsePayload));
 return;
}

The Response class is a wrapper around a SForceService instance which returns a String we can use as a logging statement.

class Response
{
    private SforceService service;
    public Response() { }
    public String Send(String objectName, Hashtable payload)
    {
        return this.Send(objectName, payload, “External_Id__c”);
    }
    public String Send(String objectName, Hashtable payload, String syncFieldName)
    {
        if (payload == null || payload.Count == 0) return

            “No records updated. Response.Send() not needed.”;
        try
        {
            service = new SforceService();
            LoginResult loginResult = service.login(
                ConfigurationManager.AppSettings[“Salesforce.Username”],
                ConfigurationManager.AppSettings[“Salesforce.Password”] +
                ConfigurationManager.AppSettings[“Salesforce.Token”]
            );
            service.SessionHeaderValue = new SessionHeader();
            service.SessionHeaderValue.sessionId = loginResult.sessionId;
            service.Url = loginResult.serverUrl;
            GetUserInfoResult userInfo = loginResult.userInfo;
            List sObjects = new List();
            sObject s = new sObject();
            XmlDocument xmlDocument = new XmlDocument();
            List elements;
            XmlElement element;
            foreach (String sfid in payload.Keys)
            {
                s = new sObject();
                s.Id = sfid;
                s.type = objectName;
                elements = newList();
                element = xmlDocument.CreateElement(syncFieldName);
                element.InnerText = ht[sfid].ToString();
                elements.Add(element);
                s.Any = elements.ToArray();
                sObjects.Add(s);
            }
            SaveResult[] saveResults = service.update(sObjects.ToArray());
            String output = String.Empty;
            foreach (SaveResult sr in saveResults)
            {
                if (sr.success)
                {
                    output += String.Format(

“{0} Response Sync Success : sf.id=\”{1}\””, objectName, sr.id) + “\n”;
                }
                else
                {
                    foreach (Error err in sr.errors)
                    {
                        output += String.Format(

“{0} Response Sync Error : sf.id=\”{1}\”, {2}”, objectName, sr.id, err.message) +

“\n”;
                    }
                }
            }
            return output;
        }
        catch (Exception ex)
        {
            return ex.Message + “ — “ + ex.StackTrace;
        }
    }
}

The credentials and API service endpoint are declared in the Web.Config file.

https://test.salesforce.com/services/Soap/u/24.0

(In a production environment, replace “test.salesforce.com” with “www.salesforce.com".)

Some caveats are

(1) The response needs credentials to login, and you might need to setup a system account for the response.

(2) To keep the message from recursing, the Response User needs to have outbound messaging disabled, or be specifically excluded from the Outbound Messaging Workflow.

(3) If your listener implements a producer/consumer queue, the response may not be immediate. (If tens of thousands of records are being updated by a slow system, it could be an hour or more!).

(4) An outbound message response may update the record while it’s being viewed by another user. If the user makes another change without refreshing the record, Salesforce will decline the second change, and the human user will have to refresh and repeat.

By allowing a response to an outbound message, Salesforce gives us the opportunity to close the loop. For example, if a record is being inserted into an external system, the response could return the external ID, for use with future updates.

While not a “point-and-click” product, like the Apex Dataloader, the outbound messaging API provides developers all the hooks we need to create a full-service data connection.