A Set of Short Talks

In this month's meeting, we had 4 members give short presentations on different topics of interest to them. Because of the number of topics and the discussions each triggered, this meeting actually ran to almost a full two hours.

Object vs Data Driven References

Robert Stone began the discussion by asking for the group's opinion on two different approaches to referencing one object from another. One approach references the object directly, the other stores an ID that is used to look up the dependent object separately. From the beginning of the presentation, most of the group seemed quite comfortable with direct reference, since this is how objects are traditionally handled. Many of us struggled to understand why one would use the new ID-based approach that Robert was proposing.

Eventually, Robert explained that he had always used the latter approach based on his database-heavy background. He was trying to understand the advantages that the rest of us saw in the Object approach.

As usual, challenging long-held beliefs one the way the world works is a great way to force you to re-consider what is obvious. This was quite a thought-provoking discussion.

Lock-Free Architectures

Julian Brown moved us from the high level of Object Design down to the level of OS algorithms by talking about blocking and non-blocking algorithms and data structures. In particular, he talked about a set of lock-free structures that he has been studying lately, endeavoring to help us understand how these structures work.

He showed how many of these structures depend on Compare and Swap (CAS) instructions that can be guaranteed to be atomic by the CPU. He also touched on Read Copy Update algorithms, which he hopes to explore at another time.

Julian's presentation is available.

More Movable Type SQL Injections

John "JD" Lightsey performed another entertaining dive into the Moveable Type source to show another SQL injection. This time the flaw was in the XMLRPC code. In particular, the code uses the symbol tables of various modules as dispatch tables for execution of code supplied in input requests. JD showed how there was no protection for private functions of these module, and fun ensues.

Once again, JD showed the power and ease of use of the sqlmap program for exploiting these SQL injections on his own servers to give and idea of how powerful they could be.

Long-time Perl Vulnerability

Todd Rinaldo discussed a theoretical vulnerability in the way the default Perl build sets up the @INC search path. To be precise, the search path contains the current directory as the last place to search for a module. Although this has been known as a theoretical problem for quite some time, the maintainers of Perl have not been willing to change that because it might break code that relies on this behavior.

In recent months, the team at cPanel and others have shown ways to exploit this vulnerability in a production system. The details are not being released until appropriate action has been taken. Todd walked through how this exploit would work and what could actually be done about it.

We had 13 people attending this month. As always, we'd like to thank HostGator, LLC for providing the meeting space and food for the group.