Generally, it is better to link to the abstract than to the PDF. If the paper gets revised the abstract page page's PDF link will point to the latest version of the paper (with links available to prior revisions). The direct PDF link will remain stuck with the old version.
Because I the find abstracts of readable, clear, and interesting papers are often indistinguishable from unreadable and dull papers, I prefer the paper itself to its abstract...at least in terms of Hacker News...I am long since reading scholarly papers for academic purposes. If someone wants to read the abstract, it's right there at the beginning of the paper anyway.
Abstracts are short enough that they can be inlined into a hackernews comment directly:
> A critical analysis of the feasibility of reversible computing is performed. The key question is: Is it possible to build a completely reversible computer? A closer look into the internal aspects of the reversible computing as well as the external constraints such as the second law of thermodynamics has demonstrated that several difficulties would have to be solved before reversible computer is being built. It is shown that a conventional reversible computer would require energy for setting up the reversible inputs from irreversible signals, for the reading out of the reversible outputs, for the transport of the information between logic elements and finally for the control signals that will require more energy dissipating into the environment. A loose bound on the minimum amount of energy required to be dissipated during the physical implementation of a reversible computer is obtained and a generalization of the principles for reversible computing is provided.
It can save readers a click in deciding if they want to go for the whole paper or not. Think of the abstract as an upper bound on whether the paper is interesting or not. A paper could possibly be less interesting than its abstract, it is much less likely that it is more interesting than its abstract unless the author really messed up.
Personally, I usually click on links based on the title more often than based on the comments. But I know other people have different habits. With a direct link to the PDF, the abstract is at the top of the PDF for anyone who wants to read it.
For me personally, if I have the paper, I skip the abstract and read the introduction because it is more interesting and I don't feel a duty to bore myself reading the abstract. Here, I thought the title was interesting enough to click on the link and the paper interesting enough to provide a link to the less noisy original.
There's a small tradeoff in that not all readers might want to commit up-front to downloading a PDF. In the case of arXiv the abstract will be also updated with later versions of the paper if they appear.
With Firefox the resource located at the arXiv URL of the PDF is smaller than the the resource located at the arXiv URL for the abstract. So goes the modern web. Both are an order of magnitude smaller than the originally submitted link.
For me, the conventionality of UI of the PDF resource makes it easier to navigate than the UI of the resource for its metadata. Admittedly, I don't have a use case that is dominated by revisions to preprint journal articles...I just want to check papers out.
Generally, it is better to link to the abstract than to the PDF. If the paper gets revised the abstract page page's PDF link will point to the latest version of the paper (with links available to prior revisions). The direct PDF link will remain stuck with the old version.