Is LTI the wrong platform for New Quizzes?

[SomeHandyGuy]

Philosophically speaking, that is. LTI seems to have been a bad choice of platform for New Quizzes as most of the complaints and feedback seems to stem from the fact that they're LTI-based. This causes me to worry because it means a lot of problems will always remain unfixable. Worse, moving to a different technology platform would mean starting from scratch.

The problem with LTI is that it means NQ are actually third-party applications that are dressed up to look like they are integrated with Canvas. They will never have the first-party access or privileges that Classic quizzes have. Part of the reason NQ are slow and can't do things is because they constantly have to make server requests for permission/access to the data they need to run. This is also the reason NQ never seems to know which course it is embedded in. It's not, it's on a completely different system and your course only links to it.  LTI does not grant full-access to student/intuitional data. It's designed for third parties and only gives out data on a need-to-know basis. This just seems unnecessary for a product made for Canvas by Canvas devs.

What does this mean for users? I think it extremely unlikely that NQ will go back to the drawing board, even if a majority of customers are screaming at them to do so. That only leaves the option of creating a New Quiz 2 or fixing Classic. Honestly, I don't understand why they just can't fix Classic? Is it really that bad? (I guess it must be.) For me, there is only one reason I can't use Classic (randomise answers at the question level) and if they fixed that I could turn my back on NQ and, for the most part, be better for it. Classic still looks and operates well and in many cases, better than NQ. Don't even get me started on that fact that now that we've invested a lot of materials to NQ and there is STILL no way to get it out. We're stuck with it.

[danaleeling]

I have wondered whether, like the spinning out of Canvas badges as Badgr, the original impetus was to create an LTI tool that could then be marketed and sold independent of Canvas. This would also allow new quizzes to be potentially be monetized even on the home platform. I do not know, but I would wonder if implementation as an LTI would also avoid some of the open source licensing obligations in Canvas core. As was noted by the original post, apparently an LTI tool does not know what course called the tool, only what user called the tool, hence the change in the way quiz banks are owned under NQ. At this point Instructure has significant sunk costs in New Quizzes. I continue to recommend Classic Quizzes and hope that by 2024 the many issues will find a resolution.

[hesspe]

I find the topic of why New Quizzes was developed as an LTI fascinating.  I heard Jason Sparks, the first Product Manager for new quizzes (or Quizzes.Next, as it was then),  present on it in June 2016 at a Canvas event near Boston, and  heard him again at InstructueCon in August 2017.  (you can still see that talk here).  Jason had a great sales pitch, and I was expecting that we might see New Quizzes replace the existing quiz application maybe by 2018, latest  Well, Jason left Instructure in August 2018.  Without any knowledge of the matter at all, I always blamed Jason for selling Instructure on doing New Quizzes/Quizzes.Next as an LTI, and then baling when it became clear that was a very bad idea.  Whether or not it might be true that at one point Instructure hoped to develop a stand alone quiz engine that they could peddle to other "Tool Users" (platforms), I think that they early on embraced the LTI concept as the wave of the future.  They were not entirely wrong in that (as I think history has shown) but where they were wrong I think is that core services such as Assignments, Discussions, Grades and Quizzes need to be able to communicate with each other and with the hub services in ways that LTI doesn't allow.  I'm sure there were people at Instructure who realized this early on, and I picture fierce arguments about it before (and perhaps even after) the decision was made, but as soon as Quizzes.Next was released to users, and people started using it, the die was cast.  The only option then would have been to support two quiz engines and Instructure, for very understandable (but to my mind misguided) reasons, has steadfastly resisted doing that.