So today I thought I’d give Proteus2 a try. Proteus2 is a protein structure prediction server. There are many around, of varying quality. This one has good pedigree, which is part of what attracted me. What I am left with is a frustration that is not new. When structure prediction servers became common, I could understand the need to have queues and all the static pages with automatic refresh. But that was then. In this day and age, things should be different, but they aren’t. The screenshot below was taken close to 2 hours after my job was submitted (it’s 4 hours as I write this, but still waiting).
In a day when we expect our search queries to be instantaneous, why can’t we expect a structure prediction server to start a job quickly. It’s not like they are all getting hammered by traffic. The reason itself is hardly surprising. The funding models, the way the services are set up, etc haven’t really changed all that much in the past 5-6 years, but our expectations for web design and what we want from a web service have. Those of you applying for funding and setting up web servers probably know this better than I do. What’s keeping academics from setting up servers on the cloud other than funding models?
Another pet peeve, not exactly related to this one, software that needs to be download (source) from a particular site. In a day with good code repositories available, I wish most projects would be hosted there, with version control and the potential of others getting involved.
Regarding the source code repository, I think Linux distributions provide good options here. Ensuring that your software is available in the Debian/RedHat/etc repository provides an excellent way to ensure your software is available, portable, easy to install etc. I strongly recommend this to any open source science project.
I agree. Basically the point is that we should use commonly available mechanisms to distribute code, rather than put them into this ether.
harijay
What’s keeping academics from setting up servers on the cloud other than funding models? I dont think the cloud is the answer to having software work well on the web. It all maintenance. Most academic projects run out of funding , or the grad student/postdoc maintainer leaves the lab. As a result you have broken web pages and stalled web jobs!
That's worth a rant in itself. Funding models are critical to take the capital costs and hardware maintenance issues out of the picture. The other issue is a more fundamental one and it's why I don't think academics should keep software locked up (even it if is open source in theory), but make it available via sourceforge, google code, etc. Still doesn't solve the problem, but it's a start.
Thing is, most academic software is badly written (I've worked with enough to know by now), so what's the solution? People don't want to pay for software, and web services are not something you can download and deploy most of the time. Is the solution to hire young programmers? The turnover rate will be too high.
Or perhaps informatics types will learn REAL software engineering.
Another wish for life science APIs and web services
In a day when we expect our search queries to be instantaneous, why can’t we expect a structure prediction server to start a job quickly. It’s not like they are all getting hammered by traffic. The reason itself is hardly surprising. The funding models, the way the services are set up, etc haven’t really changed all that much in the past 5-6 years, but our expectations for web design and what we want from a web service have. Those of you applying for funding and setting up web servers probably know this better than I do. What’s keeping academics from setting up servers on the cloud other than funding models?
Another pet peeve, not exactly related to this one, software that needs to be download (source) from a particular site. In a day with good code repositories available, I wish most projects would be hosted there, with version control and the potential of others getting involved.
It’s late. Enough ranting.
Image via Wikipedia
Technorati Tags: Proteus2, Protein Structure Prediction, Web Services