I'm not sure what is the correct term in English, but by 'complete set' I mean that the span (of finite combinations) of the set is dense in the space.
I have to show for which p ($ 1\leq p <\infty $) the following is a complete set in the sequence space $ l_p $ : $$\{ x_1=(1,-1,0,0,...) \\ x_2=(0,1,-1,0,0,...) \\ ...\\ x_n=(0,...,0,1,-1,0,...)\\ ...\}$$ (for $x_n$ the nth index is 1, followed by -1).
It is easy to show that it is true for p=2 by showing that the only sequence normal to each of these vectors is the zero sequence.
I have tried proving that it is not true for p other than 2 by showing that $\|e_1-\Sigma c_kx_k\|_p$ is always greater than 1, for the unit vector $e_1=(1,0,0,...)$ and any constants $c_k$. i didn't manage to prove it, and I can't see why it would be different for p=2, which might say that my assumption is wrong.
I have seen this question with slight variations, in which the set becomes for example $x_n=(0,...,0,1,\sqrt{1+1/n},0,...)$.
A different way for $p>1$:
It's not to hard to show the linear span of the $x_i$ contains the vectors $$y_n=(1,\underbrace{-1/n,-1/n,\ldots,-1/n}_{n\text{-terms}},0,\ldots).$$
Since, for $p>1$, we have $\Vert e_1-y_n\Vert_p =1/n^{1-1/p}\rightarrow0$ as $n\rightarrow\infty$, $e_1$ is in the closed linear span of the $x_i$. But then every $e_i$ is in the closed linear span of the $x_i$.
It follows that $\{x_i \mid i\in\Bbb N\}$ is dense in $\ell_p$, for $p>1$.
For $p=1$, see my comment above.