[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [sarif] Change bars for Issue #158 (result.correlationId)
Michael – please read this carefully to see if I’ve correctly conveyed your philosophy for analysis tool design.
1. First of all, don’t worry, you can indeed populate result.fingerprints. “SHOULD NOT” means “don’t do it unless you have a good reason,” and I guess your tool has a good reason 😊
The reason we said “SHOULD NOT” is related to a tool design philosophy that Michael has advocated. I’m not sure how explicitly we have discussed it. We want to make it as easy as possible for vendors to build SARIF-compliant tools. So we distinguish between properties that the analysis tool is “uniquely qualified” (my words) to fill, and properties that post-processors and the result management system (RMS) can fill, because they are not tool-specific. An example of a “non-tool-specific” property is run.baselineInstanceGuid. We wouldn’t want every tool to have to discover the baseline run, or to provide a command line argument to specify it.
Michael and I considered result.fingerprints to fall into the “non-tool-specific” category as well. The RMS can compute the fingerprint in the same way for all tools. A tool that has special fingerprinting requirements can use result.partialFingerprints to convey that information to the RMS.
The question as I see it is how to tell a tool vendor “Don’t worry! Someone else can fill out this field, so you don’t need to bother.” Perhaps saying that the tool SHOULD NOT fill the field isn’t the best way to convey that.
2. Multiple rule ids: I’d like to know more about this. It’s possible that we don’t mean the same thing by the word “rule”. SARIF’s Terminology section defines it as a “specific criterion for correctness verified by a static analysis tool.”
So from SARIF’s point of view, if a programming artifact violates more than one “criterion for correctness,” that means they are two distinct results. One criterion for correctness might be “Don’t use weak hash algorithms.” Another might be “Don’t use tainted data without sanitizing it”. SARIF wouldn’t put both of those in the same result, even if the same simulated execution trace (code flow) revealed both problems.
Could you give me an example of a single result (or “finding,” or “observation”) made by Fortify that violates more than one criterion for correctness?
Sorry for not bringing these up earlier, but I have a couple of comments:
Normally I incorporate amendments adopted by the TC into the provisional draft without asking for further review. In the case of Issue #158 (Introduce result.correlationId and clarify purpose of result.fingerprints array), the changes were substantive enough that I wanted to show them to you explicitly. I’ve attached a change-barred version of the provisional draft that shows the changes I made based on the TC’s feedback.
I am going to merge these changes, along with the other changes we adopted today. But if you disagree with the way I incorporated the feedback on #158, now’s your chance to tell me.